Anal. Chem. 1996, 68, 21R-61R
Chemometrics Steven D. Brown,* Stephen T. Sum, and Frederic Despagne†
Department of Chemistry and Biochemistry, University of Delaware, Newark, Delaware 19716-2522 Barry K. Lavine
Department of Chemistry, Clarkson University, Potsdam, New York 13676 Review Contents Software Books Tutorials Statistics Optimization Signal Processing Resolution Calibration Parameter Estimation Structure/Activity Relationships Pattern Recognition Library Searching Artificial Intelligence Literature Cited
23R 23R 25R 25R 28R 30R 35R 36R 41R 42R 44R 47R 47R 50R
Chemometrics is the discipline concerned with the application of statistical and mathematical methods, as well as those methods based on mathematical logic, to chemistry. This review, the eleventh of the series, and the ninth with the title “Chemometrics”, covers the more significant developments in the field from December 1993 to November 1995. The format follows that of the previous review (A1) of this subject. Not surprisingly, the number of citations of chemometrics in general continues to show steady growth in all areas. Publications concerned with development of new chemometric methods showed only slight growth during this period, as measured by the number of citations, no doubt because the number of chemometrics researchers actively engaged in development and publication of novel chemometrics methods has remained about constant. On the other hand, the number of those applying chemometrics continues to grow, and the number of publications concerned with applications of chemometric methods also grew substantially. Approximately 25 000 computer-generated citations were screened for this review. A hand search located many more citations missed in the computer search. In view of the need to make this review as concise as possible, we have pared the large number of citations, many of which report routine application of basic chemometric methods, down to about 1200 for mention or brief comment here. Facets contributing to growth in chemometrics included the continuing increase in chemometrics by those in statistics and mathematics, the slow increase in papers submitted by workers in chemical engineering, and the growth of chemometrics workers in several areas, including Europe as well as Asia. With the steady increase in papers from China and Eastern Europe, there were a few surprises in the geographical distribution of papers dealing † Present address: Farmaceutische Instituut, Vrije Universiteit Brussel, Laarbeeklaan 103, B-1090 Brussels, Belgium.
S0003-2700(96)00005-4 CCC: $25.00
© 1996 American Chemical Society
with chemometrics. One was the large increase in papers from Spain. There are now several research groups active in that country, and their productivity has been quite remarkable. A second was a substantial increase in papers published by British authors. As in Spain, chemometrics seems to be expanding rapidly in Britain. Probably the most noteworthy item that occurred during the last two years has been the influence of the World-Wide Web on the chemometrics community. The Web-based chemometrics virtual conference, The First International Chemometrics InterNet Conference, or INCINC-94, was held during September-December 1994. The INCINC-94 papers presented at the conference can be found at the location http://emsl.pnl.gov:2080/docs/ incinc/. This meeting, one of the first two chemical conferences of any sort held on the Web, put the field at the forefront of scientific disciplines making use of the Web for exchange of ideas. The format of this conference offered a number of advantages for presentation of chemometrics material. One was that the audience could control the rate of delivery of information. Chemometrics is a subject that can be difficult for the audience to assimilate at the rate needed for the author to give background, new material, and results of applications all in a typical 35-min presentation. The presentations given at the INCINC-94 conference were static rather than dynamic, in that they could be downloaded as text, with supporting figures and tables, and examined leisurely, rather than assimilated from a verbal stream given at a pace forced by a timed presentation. Having the presentation in a fully fleshed-out manuscript form gave the audience an opportunity to look at each presentation more carefully. Comments and questionssand their responsesscould be sent during a two-week period rather than the usual few minutes (or less) following the presentation. Another advantage was the possibility of getting well-known speakers from all over the world without need to solicit travel funds for the conference from increasingly reluctant funding agencies. The audience was able to attend the presentations without the need to travel or even to leave work, a mixed benefit, but one that probably raised the attendance. There were a number of difficulties raised by the Web conference format, however. Having manuscript-like “talks” available to the public would seem to blur the line between presentation and publication. The format is similar to the first days of published scientific research, where papers were “read” and discussed at gatherings of members of some scientific society, and were subsequently printed as a convenience for those unable to attend those gatherings. Indeed, the presentations at INCINC94 were referenced as publications by Chemical Abstracts. The publication/presentation distinction is further complicated by the fact that many of the presentations were subsequently published Analytical Chemistry, Vol. 68, No. 12, June 15, 1996 21R
in an issue of Chemometrics and Intelligent Laboratory Systems that was devoted to this conference. Otherwise, the impact of the Web on chemometrics is just beginning to be felt at the time that this is written (January 1996). Several research groups have established Web home pages with information on publications and on-going research. Four of the vendors of commercial chemometrics software also have Web home pages. A recent Web-search showed some 238 hits on the terms “chemometric” and “chemometrics”. For the most part, however, the home pages for chemometrics sites just offer links to each other, or occasionally to home pages outside of chemometrics, such as home pages for Statlib, the PNL neural networks information gateway, or the MathWorks home page. In addition to the Web site holding papers from INCINC-94, there is one Web page with an electronic course on chemometrics (http://odin.chemistry.uakron.edu/chemometrics/), and one with an electronic database of chemometric publications (http://newton.foodsci.kvl.dk/chemobro.html). A few chemometrics web sites offer software. Projects such as the chemometrics data set repository appear to be moving slowly, and sharing of information among chemometrics groups has not changed much with the advent of the Web. Yet, the Web is changing and growing daily, and it is hard to predict what will be the state of chemometrics on the Web in the five months between when this is written and when it will appear in print, let alone a year. The growth of the Web and the spreading of chemometrics to a wide range of applications means that information about chemometrics is tending to become more diffuse. As chemometrics expands to distant fields, and more authors publish novel methods in journals focused on an application in some field distant from the core chemometrics area, there is a need to rely more heavily on electronic searches rather than a quick scan of the few core journals where chemometric articles appear. Unfortunately, the accuracy of electronic searches does not appear to be improving. As noted in previous reviews, important articles on chemometrics were not detected in the searches of Chemical Abstracts. Many of these articles appeared in the “core” chemometrics journals. As before, hand searches of the journals where chemometrics articles were most likely to appear were still necessary to find many citations. The automated searches located a good deal of work not relevant to a chemometrics review, however. In this review, about 5% of the citations appearing here resulted from hand searches of the core journals and several journals where chemometrics is known to appear fairly regularly. The Web searches also reflect this problem: many well-known chemometrics groups’ sites were undetected (though known to exist) by several search engines, while the search engines turned up a number of what could only be called obscure sites. Apparently, work remains to be done on perfecting the chemometrics database and search methods in CAS and Web searches. Keyword selection by many authors and prompt abstracting of articles appear to be the biggest obstacle to locating work on chemometrics in a CAS search and a Web search. Authors can help by thinking about searches when they select keywords, and those with chemometrically oriented Web pages can make their site more likely to be found in a search by passing their address to one or more of the Web sites that collect information and abstracts of Web pages, as well as those chemometrics sites performing indexing of relevant pages. 22R
Analytical Chemistry, Vol. 68, No. 12, June 15, 1996
Reviews appearing on aspects of chemometrics were fairly numerous. The largest and most general review of chemometrics continues to be the biannual Fundamental review (A1). Process analytical chemistry, which includes a number of chemometrics applications, is also reviewed every two years in Analytical Chemistry. The most recent review appeared in 1995 (A2). Bourguignon et al. offered a brief glance at chemometrics in their overview of the field (A3). The use of chemometric methods in generating mathematical models for chemical systems was the subject of a review by Rao et al. (A4). Franke and Gruska reviewed in detail the applications of principal components analysis and regression in medicinal chemistry (A5). Clementi et al. also reviewed the role of chemometrics in medicinal chemistry, focusing on the development of chemometric models (A6). Chemometrics in the Benelux, an area with many active research groups in academia and industry, was reviewed by Vandeginste (A7). Cheng (A8) discussed the state of analytical chemistry, including chemometrics, in China. Geladi summarized activities at Eastman Kodak in a short review (A9). Schlager and Ruchti discussed the theoretical background of chemometric methods used by their company (A10). The role of chemometric methods in lipid analysis was the subject of an overview by Kaufmann (A11). Chemometrics has long been recognized as effective means of solving many of the complex problems that arise in food research. Ashima (A12) discusses the role of chemometric methods in problems concerning nutrition, safety, and preference. Multivariate mathematical methods, including multivariate calibration and classification, and their use in pharmaceutical development, are the subjects of a review by Lindberg and Lunstedt (A13). Enviromental applications of chemometrics also attracted a good deal of interest. Wenning (A14) published a sizable review of the application of chemometrics for analysis of complex mixtures in environmental matrices. Brereton (A15) summarized chemometric methods for dealing with trace contaminants in the ecosystem in another review. Environmetric methods for dealing with trace element data were reviewed by Vogt et al. (A16). The close relationship between spectroscopic analysis and chemometrics offered material for several reviews. Sherwood offered an extensive review of data analysis methods in X-ray photoelectron spectroscopy (A17). The integration of chemometric methods into surface spectroscopies was the subject of a review by Hutter and Grasserbauer (A18). Strasters (A19) examined the use of chemometric methods for reduction of data from HPLC with diode-array detection. Mc Clure’s (A20) review of the state of near-infrared spectroscopy included a large section on chemometric methods critical to that technique’s success. The use of near-infrared spectroscopy and multivariate calibration in petrochemical processing was the subject of a short review by Descales et al. (A21). A more general review by Oshima et al. considered the role of chemometrics in practical applications of analytical spectroscopy (A22). Others concerned with the practical aspects of data analysis included Trebbia (A23), who examined the assumptions behind modern methods for data analysis in a recent review, and Converse (A24), who examined the details, both chemometric and practical, involved in at-line analysis. Miller summarized the use of chemometric methods in process analytical method development (A25). A few authors reviewed recent advances in chemometrics. Carey considered one- and twodimensional sensor arrays and the use of appropriate chemometric methods for environmental applications (A26). Henrion published
an extensive review on three-way methods in chemometrics (A27). This review summarized the theory of Tucker’s model and demonstrated three-way analysis in applications. Kraus et al. offered a brief overview of neural networks in chemical analysis (A28). To date, chemometrics has dealt with systems as deterministic or random, yet many chemical systems behave chaotically. Bishop discussed the relationship of chaos theory, chemometrics, and chemistry in an interesting review (A29). In what is probably the most provocative review of the past two years, Booksh and Kowalski considered the theoretical aspects of analytical chemistry and showed how the first- and second-order multivariate calibration methods can be considered in terms of well-worn analytical concepts such as figures of merit. The integration of a wide array of newer chemometric methods with the basic principles of quantitative analytical chemistry made for interesting reading (A30). Two other provocative papers concerned with the state of chemometrics, its future, and its relationship with closely allied fields also appeared during the past two years. While not reviews in the narrow sense, they reviewed aspects of the current state of the field of chemometrics. The paper by Wold focused on factors that distinguish chemometrics from statistics (A31), while the paper by Brown expressed concern for future growth, citing structural weaknesses in the field resulting from its historical development (A32). Both papers were part of the INCINC-94 conference. In addition to papers presented at the INCINC-94 virtual conference, many of the international and national chemometrics conference proceedings were published. The proceedings of the Third Scandinavian Symposium on Chemometrics and CAC-V appeared, to name a few. Proceedings from the 1992 Snowbird Conference appeared as a book in 1994, and those from the 1994 Snowbird Conference will appear as a book in early 1996. Several conferences devoted to chemometrics are scheduled for 1996-1997. The Chemometrics in Analytical Chemistry (CACVI) conference will be held in Tarragonia, Spain, in June 1996. The first European visit of the Gordon Research Conference on Statistics in Chemistry and Chemical Engineering will follow at Oxford, UK, in late July, and the second INCINC virtual conference (INCINC-96) will occur during the fall. Other regional conferences and more general conferences with a sizable component involving chemometrics are also scheduled. As it is now becoming common for conferences to establish their own Web pages, readers may wish to consult the Web for details and additions to the list. Software. The rate of production of user-written software for chemometrics has declined notably during the past two years, no doubt because of the quality and availability of commercial chemometrics packages. Nevertheless, a few groups continue to produce chemometrics software of high quality. The program HOLMES was reported for use in target factor analysis (A33, A34). INSPECT, a well-written general purpose program with a nice user interface for visualizing and interpreting data, was published by Lohninger (A35). Several packages appeared for use in structure/activity work. SPECTRE-M, a package for modeling of structure/activity relations using partial least-squares regression (A36), the OASIS system for structure/property relation modeling with a library of three-dimensional molecular models (A37), and ToSIM, for creation and handling of twodimensional databases of chemical structures (A38), all appeared
recently. Spreadsheet applications also offer a simple way to get data and chemometrics together without extensive programming. Two macros for Excel spreadsheet software, one implementing several methods for weighted regression (A39) and the other a method validation module for calibration (A40), were reported. The Solver tool, a macro built into Excel version 5, was evaluated for use in nonlinear curve fitting (A41). Some unusual software continues to be developed to solve special problems. Majoras et al. report an analytical engine in C++, an object-oriented language, for peak fitting across a wide range of spectroscopic techniques (A42). Two public domain software packages for image analysis on the Macintosh computer, the MacLispix program (A43) and the NIHImage program (A44), were described. Software for Monte Carlo simulation of transmitted, backscattered, and secondary electron signals from scanning electron microscopy (SEM), for use in interpreting line widths of measured SEM signals as well as image analysis, has also been described (A45). The PLS Toolbox, at the time a public domain package for multivariate calibration and related chemometric methods, but now a commercial product, was demonstrated in process analytical chemistry (A46). Finally, the DETARCHI program for detection limits with specified assurance probabilities and characteristic curves was published (A47). An issue raised that concerns software also merits mention here. Phillips (A48) expressed concern with the validation of algorithms and other software imbedded in analytical instrumentation. As commercial chemometric software becomes more sophisticated, users need assurance that the software is behaving as expected. A few vendors of commercial, stand-alone chemometric packages have taken steps to provide this assurance. Their software has been validated against other, established software whose algorithms can be inspected and tested for accuracy in the range of applications expected by the user. Some commercial, stand-alone chemometrics packages have been validated by U.S. FDA protocols. Much of the imbedded chemometric software sold now has not been validated externally, however. The prudent user of chemometrics is wise to do what is possible to extend the quality assurance testing of new instrumentation to the data analysis and to demand from vendors at the point of purchase the capability of performing validation of the data analysis imbedded with analytical instrumentation. Books. The past two years saw continued activity in the preparation of new texts on chemometrics and on more general texts concerned with special techniques in data analysis. Notable additions to the chemometrics library included a handbook of data analysis by Frank and Todeschini (A49). Volume 2 of the series Chemometrics for Analytical Chemistry, by Melloun, Millitky, and Forina, also appeared in the last two years (A50). This volume focused on principal component-aided regression and related methods. A festschrift in honor of Prof. Gerrit Kateman, who retired in 1994, was published with chapters containing contributions from many of his co-workers in Nijmegen (A51). Structure/ activity and structure/property relations is an area with rapidly increasing interest from fields outside of chemometrics. Collier edited a text covering two-dimensional and three-dimensional searches and modeling as well as neural network applications to structure/activity relations (A52). van Waterbeemd edited a nice compilation of contributionssfrom the basics to more advanced topicsson structure/property relations (A53). Spectroscopic applications of chemometrics were the subject of several books. Analytical Chemistry, Vol. 68, No. 12, June 15, 1996
Adams offered a tutorial guide to the use of chemometric methods in analytical spectroscopy, including pattern recognition and calibration (A54). George and Steele revised their 1990 text covering computer methods in molecular spectroscopy to include more on chemometric methods (A55). A book on the proceedings of the 1992 Snowbird Conference on Computer Enhanced Analytical Spectroscopy, which includes many chapters dealing with advances in chemometric methods, also appeared (A56). Many new books on software and languages for data analysis will be of special interest to workers in chemometrics. Scientific computation, an area that is changing rapidly with the creation of next-generation languages and new computational hardware, was the subject of a new text by Bellomo and Preziosi (A57). This text focused on modeling methods in the applied sciences. One of the next-generation languages familiar to many in chemometrics is MATLAB. Sigmon’s MATLAB primer, long available for early versions of the MATLAB language by ftp, has been published for MATLAB version 4 in book form (A58). S and its close relative S-PLUS are two more of the newer languages occasionally used by those in chemometrics and by many in applied statistics. Several new texts on S and S-Plus appeared. Marazzi offered a text on the use of S for robust statistical methods, and he provided many algorithms for practical use (A59). Everitt’s handbook of statistical methods makes use of S-PLUS for many of the standard statistical methods provided in SAS or other stand-alone statistics packages (A60). Venables and Ripley have published their longawaited text on the basics of S-PLUS and the practice of modern statistics using the S-PLUS software (A61). This text comes with extensive data sets for self-instruction. XploRe is another new language for data analysis focusing on graphics. The new text by Ha¨rdle et al. illustrates the use of XploRe on subjects in applied statistics and econometrics, many of which are familiar to those in chemometrics: projection pursuit analysis, and density and regression smoothing, to name just a few (A62). A number of books were published on the use of Mathematica in the sciences. One that may be of interest shows creation of decision support tools using Mathematica as a means for generating statistical decisions on such things as fractal time series and chaotic systems (A63). The C language, while older than the specialized languages mentioned above, is still heavily used in the sciences. A book describing a new numerical library in C will no doubt appeal to many chemometricians who program in C (A64). New texts focused more on data analysis and less on languages appeared on a number of subjects of interest to those involved in chemometrics research. Wavelets are an active area of research in many fields, including chemometrics. A new, introductory text on wavelet theory and practice will certainly be useful to workers in this new area of research (A65). Cleveland offered a revised version of his book on graphing data (A66). A new text on practical application of statistics in engineering covered several subjects pertinent to chemometrics (A67). The principles of experimental design were covered in a revised versions of Kempthorne’s text (A68). Several texts updated instruction on statistical methods for process control and quality control. Methods for statistical process control and total quality management were covered in another new text (A69). Quality assurance and the statistical methods to perform quality control were covered in a text by Mittag and Rinne (A70). Mandel, well-known to chemometricians for his original thinking, published a monograph on two-way layouts and their application in several scientific 24R
Analytical Chemistry, Vol. 68, No. 12, June 15, 1996
disciplines (A71). Mathematical modeling and multivariate analysis were also given fresh examinations by several authors. Mesterson-Gibbons (A72) focused on the conversion of physical problems to mathematical models in a new text. Birkes and Dodge published a text on alternative methods of regression, with coverage of robust regression, Bayesian regression, and ridge regression (A73). Rencher’s text introduced multivariate analysis with applications and featured SAS command files for the student (A74). Everitt and Dunn discussed multivariate analysis of real data in their new text (A75). The third edition of Barnett and Lewis’s text on outlier methodology appeared (A76). Optimization methods have also benefited from exploration in several new texts. Chong and Zak offered an introduction to constrained optimization of linear and nonlinear systems of equations (A77). Myers and Montgomery considered response surface methodology from the perspective of experimental design, covering mixture designs and robust designs (A78). Shao and Tu (A79) examined the theoretical basis and applications of the jackknife and bootstrap, two computer-intensive methods which are finding increasing use in chemometric applications. The bootstrap, cross-validation, and validation are also covered in a more introductory text by Hjorth (A80). Other texts covered a wide range of methods with some relevance to chemometrics. Discriminant analysis is the subject of a new book by Huberty (A81), who provided disks with data sets and software to illustrate methods covered in the book. Software for pattern recognition is also the emphasis of another text on pattern recognition by Weizman (A82). A new edition of Fuller’s book on statistical time series analysis covers new developments in time series analysis, including statistical treatments of the Kalman filter (A83). Artificial intelligence continues to be a popular area for new texts. Hand edited a collection of papers on the frontiers of artificial intelligence in statistics (A84). Many texts concerned neural networks and their applications in the sciences and engineering. Bulsari edited a text covering the basics of neural networks for chemical engineers (A85). Pham and Xing considered the use of neural networks for prediction and control in their text (A86). Bishop evaluated neural networks for pattern recognition (A87). Several books dealt with implementation of neural networks in the C++ object-oriented programming language. Masters discussed several unusual algorithms for neural networks from the perspective of C++ coding and included software on an accompanying disk (A88). Weistead examined neural network and fuzzy logic techniques implemented in C++. The accompanying disk provides code and a user interface (A89). Pattern recognition with neural networks in C++ was the subject of a book by Pandya and Macy (A90). With the large amount of C++ code appearing for neural networks, it is nice to know that a new primer on C++ aimed at scientists and engineers has appeared (A91). And, for those wanting a text to help them learn about the Web, the internet, and related subjects, material that very rapidly becomes dated, a new 1995 edition (A92) of the 1994 text by Gilster is worth a look. Books are becoming very sophisticated. It has now become common for authors to bundle software with their texts. An increasing reliance on multimedia by authors is to be expected in the future, given the increasing availability of hardware and software supporting sound and video. For the moment, diversification of instructional methods seems to be on a slow increase in chemometrics. For example, Massart and Lewi have come at
multimedia instruction from a traditional route: their tutorial material on principal components analysis has been put together as an audiovisual course, with videotapes, a manual, and an accompanying software package including Lewi’s SPECTRAMAP software and MATLAB code for principal components analysis (A93). Tutorials. The current trend to cross-disciplinary thinking, of which chemometrics is a part, often involves a multivariate approach. As interest in methods for quality assurance, multivariate calibration, and structure/property relations continues to grow outside of chemometrics, there is a need for educating scientists with a wide range of skills and backgrounds in the basics of chemometric methods and thought. At present, chemometrics short courses and published tutorial material continue to serve as the main instructional aids for the beginner. The journals Analyst and Chemometrics and Intelligent Laboratory Systems continue to offer tutorial articles on a fairly regular basis. A few other journals, such as The Journal of Chemometrics and Analytical Chemistry, offer tutorials less regularly. Tutorial courses offer a depth and breadth unmatched by the typical tutorial article, providing a fast track to those motivated scientists beginning in chemometrics. It therefore comes as no surprise that chemometrics short courses seems to be increasing in number and frequency. Courses range from a two-day course in multivariate calibration and classification offered by the American Chemical Society at the Pittsburgh Conference and elsewhere to the weeklong course on offered yearly at the University of Bristol. The issue of course content in chemometrics was a significantsand popularscomponent of the recent INCINC-94 conference. INCINC94 also involved discussion of the published tutorial format by Workman, who, with Mark, continue to provide a valuable service to the chemometrics community with their serialized tutorials in Spectroscopy. The year 1994 saw them conclude their tutorial series on elementary matrix algebra (A94) and begin one on experimental design (A95). Meloun and Millitky continued their own tutorial series with a paper examining several data transformations (A96). MATLAB, as noted above, is the software language of choice of many in chemometrics. Two recent tutorials show how MATLAB can be used in chemometrics applications (A97, A98). A few other software languages also have their advocates. Brereton published a tutorial review on the use of object-oriented programming (A99), and Kragten (A100) reviewed spreadsheet software methods for calculating confidence intervals and standard deviations. Brereton introduced methods for resolution of mixtures using principal components analysis (A101). Several authors covered a wide range of multivariate methods in tutorial articles. Kourti and MacGregor introduced multivariate methods for process analysis and control in their tutorial (A102). Thomas offered a well-written primer on the statistical underpinnings of multivariate calibration (A103). Brown summarized the use of latent variables in multivariate calibration and classification in another tutorial article (A104). Ho¨skuldsson illustrated the H-principle in multivariate calibration and provided algorithms as well as a number of examples in his tutorial (A105). Finally, members of the Kateman group offered a two-part introduction to neural networks of various types (A106, A107). STATISTICS The number of publications concerning statistics in the last two years remained more or less constant relative to the previous
review period. That is, roughly as many papers whose primary focus was statistical methodology were reported. For reasons of scope, the articles cited here omit those containing merely the application of established statistical methods. Furthermore, this section excludes research outside the realm of analytical chemistry. However, many citations in the statistics journals such as Technometrics and the Journal of the American Statistical Association may be of interest. In the chemistry literature, statistics was the subject of research for numerous reasons such as interlaboratory comparison, method validation, determination of detection limits, development of sampling strategies, error analysis, process control, and outlier detection. A number of reviews on statistics were published during the last two years. For instance, some fundamentals of statistics for chromatographers were given (B1). In addition, Howarth provided an overview of quality control statistics for the analytical laboratory. He discussed many univariate techniques including various types of charts (Shewhart, X, R, etc.), robust statistics, outlier detection, boxplots, and use of control materials (B2). Maier and co-workers discussed the many purposes of interlaboratory studies such as proficiency testing, learning exercises, quality control, and certification (B3). In a four-part review, statistical analysis of chemical data was described using a variety of standard methods. Among the main topics, exploratory data analysis, analysis of sample distributions, data transformations, and parameter estimation were discussed (B4-B7). In another review, Danzer explained the statistics of quality assurance (B8). The assessment of radiological data using various parametric and nonparametric methods was also described (B9). Method validation was the topic of several reviews. For example, Wegscheider provided a general overview of statistical tools for the validation of analytical methods (B10), while Grimstvedt gave one in the context of geochemistry (B11), and Rymen furnished one regarding the determination of dioxins (B12). Two reviews were presented on the error analysis of analytical procedures. In one, the detection and elimination of systematic errors in trace analysis was described with an extensive list of references (B13). In another, measurement uncertainty and its use in the interpretation of data quality was reviewed (B14). Robust statistics were also discussed by chemistry researchers. Xie et al. explained the basics of robust statistics including their application to multivariate data (B15). In addition, an overview of robust statistics for process design was provided by Grize (B16). Several papers regarding the use of statistics for assessing the performance of laboratories in interlaboratory testing appeared in the literature. In one of these, Heydorn described the detection of systematic errors in analytical results from different laboratories (B17). In a two-part report, Mudd and co-workers discussed the interlaboratory comparison of autoradiographic DNA profiling measurements. Part one described the data and provided some summary statistics (B18), while part two developed a model for band-sizing variability between laboratories (B19). Another report described the QUASIMEME interlaboratory performance studies. Two measures, one for bias and another for precision of laboratories, were given, and practical considerations of robust statistics for interlaboratory evaluations were addressed (B20). A methodology for testing clinical laboratories in the analysis of specimens differing in analyte content was proposed. Concentrationdependent profiles were presented as versatile means for depicting interlaboratory scatter (B21). In another study, analysis of Analytical Chemistry, Vol. 68, No. 12, June 15, 1996
covariance was introduced for the simultaneous evaluation of concentration-independent and -dependent data from different laboratories. The study showed that analysis of covariance could better detect the presence of error than the usual analysis of variance (ANOVA) method and that the former was effective at isolating the different sources of error (B22). Many companies world-wide employ chemical measures for quality control. Feinberg addressed the issue of evaluating the quality of analytical methods by discussing several facets of statistical analysis (B23). A paper regarding laboratory proficiency tests discussed the difficulties associated with establishing a reference material and determining its true value. The influence of the homogeneity of the reference material and of the algorithm employed to calculate the true value on the outcome of the proficiency test was investigated (B24). Gaskin presented a graphical method for displaying duplicate data from different laboratories using elliptical contours (B25). The total error in interlaboratory comparisons consists of several components. A strategy was given for estimating all error components in the difficult case where the errors are concentration-dependent (B26). A number of articles were published with respect to the validation of analytical methods. The International Union of Pure and Applied Chemistry (IUPAC) revised its protocol in this regard (B27). Massart et al. introduced the concept of method validation, including the statistical methods for calculating the various relevant figures of merit (B28). In an another publication, the technique of standard addition was described for validating analytical instrumentation. Functions were provided for estimating the parameters of interest (B29). In the validation of highperformance liquid chromatography (HPLC) methods, Grize and others stressed the importance of controlling the integration parameters (B30). A program based on maximum likelihood estimation (MLE) and modified elliptic joint confidence regions was also presented for method validation (B31). Ellison and coworkers described the influence of numerical precision, algorithm choice, and dynamic range on commonly computed statistical parameters and calibrations. In addition, they proposed a procedure for generating reference data sets for the purpose of evaluating statistical routines (B32). Several papers concerned specifically the validation of bioanalytical methods. For example, the Washington Conference Report on bioanalytical methods was examined by Hartman et al. They cautioned against a too literal interpretation of the requirements for accuracy and precision (B33). In another bioanalytical report, Lang and Bolton presented a procedure for pharmaceutical method validation (B34). Many aspects of bioanalytical method validation such as prerequisites, validation criteria for reliability and overall performance, revalidation, cross-validation, and others were also discussed (B35). The estimation of detection limits of analytical procedures was investigated by several researchers. Kuselman and Shenhar described the planning of experiments for determining detection limits on the basis of calibration data (B36). In a comparison study, both a calibration function and the blank deviation were utilized in establishing the limit of detection. The procedures were applied to analyses by inductively coupled plasma atomic emission spectrometry (ICP-AES) and voltammetry (B37). Eaton and Sanders investigated the feasibility of estimating practical quantitation levels based on method detection levels modified by a multiplier. They concluded that although the approach was feasible, the computation of compound-specific multipliers was 26R
Analytical Chemistry, Vol. 68, No. 12, June 15, 1996
unwieldy (B38). In another study concerning detection limits, three methods for dealing with censored data sets were compared. These were the EPA δ log-normal distribution procedure, Cohen’s MLE, and the regression of normal order scores. Cohen’s technique demonstrated superior predictive power (B39). The limit of discrimination, limit of detection, and sensitivity were the topics of discussion of another paper. Several methodologies for computing these measures were described and compared using sulfate determination in tap water for illustration (B40). Gautsch and co-workers developed a new technique for the determination of detection limits, quantification limits, and power of definition based on precision profiles and corresponding confidence intervals (B41). In a new approach, Chung presented a procedure for computing the covariance matrix of data comprising observations below multilevel detection limits. The procedure was based on marginal maximum likelihood estimation (MMLE) (B42). Various sampling strategies and ways of evaluating them were reported during this review period. Ramsey and co-workers investigated the validation of sampling procedures by collaborative trial using nested analysis of variance (B43). Further, they extended concepts from proficiency and data quality testing to the evaluation of sampling quality (B44, B45). Environmental sampling and analysis was the focus of another discussion by Ramsey, who argued for the importance of providing estimates of accuracy and precision (B46, B47). Reference limits based on reference samples that are smaller than the study sample may lead to erroneous conclusions if the typical binomial test is used. Van der Meulen et al. suggested a nonparametric test appropriate for reference samples of any size (B48). The miscellaneous schemes for sampling in spatially dependent cases were also discussed by several groups. Scott described the different statistics applicable to the detection and analysis of spatial dependence, including the trend surface analysis and kriging mapping methods, as well as approaches to spatial sampling (B49). Sampling design for measurements in soils exhibiting moisture variability was also investigated by observing the effect of sampling pattern and spatial variability on the efficiency of the sampling (B50). In a site contamination study, a method was developed for generating contamination maps based on several easily measured indicators. These maps were compressed into a single map representing site spatial variability, thereby aiding the choice of sampling location for chemical analysis (B51). Two papers discussed sampling frequency in time-dependent analyses. In one, the optimum sampling frequency based on time series statistics was determined for groundwater quality monitoring (B52). In another, parameters estimated from a time series model for atmospheric fine particle mass were used to study the influence of sampling frequency on the ability to detect trends (B53). Numerous articles were also published regarding the estimation of measurement and modeling error. Doerffel explained several statistical models useful for error estimation in analytical data (B54). Another report described improved estimation of standard deviation from possibly biased duplicate measurements. This was accomplished using trimmed plots of standard deviation estimates as a function of the quantiles of a half-normal distribution (B55). Similarly, Hueck investigated different strategies for calculating the sample standard deviation. He claimed that several commonly used methods led to biased estimates of this parameter (B56). In a different approach, standard deviations and confidence intervals were calculated using a numerical procedure (B57).
Thompson and Lowthian derived an expression similar to the function due to Horwitz for modeling the precision of proficiency tests (B58). Another strategy for estimation of measurement error using sample estimates of discrete variance function profiles was described. This method was designed to be applicable to any given protocol (B59). Cammann discussed matrix interferences and the resulting systematic errors arising in chemical sensor measurements. The conditions necessary for the proper use of correction methods were explained (B60). Statistical analysis of chemical data was covered in a report by Meloun and Militky. An analysis of indirect measurements based on Taylor series and Monte Carlo simulation was given (B61). In a two-part article, calibration in capillary electrophoresis was discussed. Part one compared the precision of peak area and peak height for carrying out a calibration (B62). Part two explained heteroscedasticity and its implication in least-squares regression (B63). Linearization of the data was shown to increase heteroscedasticity and adversely affect the least-squares model. These effects were countered by transforming the data using sensitivity weights (B64). Several expressions were derived for relating the variance of dual-energy γ radiation measurements to various system variables (B65). In another study, random effects in infrared spectroscopy were fit to models of uncertainty. The models were used in concentration estimates of the analyte (B66). The repeatability of spectrophotometric assays was investigated by comparing four statistical methods of evaluation. It was found that the standard error of the slope combined with an analysis of covariance served this purpose (B67). Propagation of error for isotope dilution analysis was utilized in the determination of fission products and actinides by ICPMS (B68). Hayashi and Matsuda studied the random error of dilution associated with the use of the volumetric pipet and flask. Mathematical expressions relating concentration error to dilution were derived using probability statistics and Monte Carlo simulation (B69). A Monte Carlo simulation was also suggested to circumvent the inversion problem in the error propagation by a covariance matrix for small-angle scattering data treatment (B70). Kvalheim et al. studied signals characterized by homoscedastic or heteroscedastic noise to determine the influence of the nature of the noise on the preprocessing technique of normalization. They concluded that heteroscedasticity had an adverse effect on the correlation structure of the data and, hence, recommended one of two transformations of the noise from heteroscedastic to homoscedastic prior to normalization (B71). Similarly, Rietjens investigated the influence of error propagation and closure due to normalization on data analysis. Two normalization methods were proposed for reducing spurious correlations due to error propagation, and a third was suggested to compensate for correlations due to closure (B72). Faber et al. derived the theory of standard errors in the eigenvalues of the covariance matrix of principal component analysis (PCA) and explained the utility of the error estimates in PCA, multiple linear regression (MLR), and the generalized rank annihilation method (GRAM) (B73). Booksh and Kowalski also investigated the influence of error on the precision of the GRAM method. Collinearity and the level and distribution of noise were discussed in this context (B74). In another report, expressions were developed for the prediction of bias and variance in the eigenvalues of GRAM and of Lorber’s rank annihilation factor analysis (RAFA). Further, a bias correction strategy was suggested for cases of small bias estimation error (B75, B76). In a two-part discussion of random error bias in PCA,
Faber and co-workers developed equations for the prediction of bias in the eigenvalues of PCA and the singular values of singular value decomposition (SVD) for the case of any number of significant principal components. They also discussed the validation and evaluation of bias predictions for a specific situation and the implication of random error bias in calibration and prediction by ordinary least-squares (OLS) regression, principal component regression (PCR), partial least-squares (PLS) regression, and GRAM (B77, B78). In another report, a correction factor for the estimate of the prediction error of individual samples was derived for PLS (B79). Two groups reported on the statistics of regression models with errors in both variables. Kalantar et al. examined the bias in linear regression parameters when both x and y were subject to either homoscedastic or heteroscedastic random error (B80). Similarly, Riu and Rius discussed calibration methods, outlier detection, and robust statistics with error present in both variables (B81). Several studies focused on the use of confidence intervals for providing error estimates. The quality of linear regression in the calibration of ICP-AES was deemed best evaluated by means of confidence limits for the concentration of the analyte (B82). Chambless and co-workers compared three methods of error propagation, namely, an approximate procedure, a Monte Carlo technique, and the “exact” method for the calculation of confidence intervals in radiation measurements (B83). Four alternatives to the maximum variance method for estimating confidence intervals in counting experiments were also discussed. These were then generalized for noncounting experiments as well (B84). In a twopart paper, Gatz and Smith first compared three methods with bootstrapping for computing the standard error of a weighted mean concentration (B85). They then compared confidence intervals calculated with the typical assumption of normality with those computed on the basis of distributions determined by bootstrapping (B86). Several papers discussed the subject of statistical process control. Veltkamp derived multivariate statistical process control models using SVD and methods of statistical process control (SPC) to monitor process changes in real time (B87). In another process control application, a neural network was trained to predict reaction outcomes based on spectroscopic information (B88). Sebastian and others applied polynomial moving-average filters (Savitsky-Golay) or data bounding to a time series model of a production process. The detection of outliers by comparison of the smoothed data with the original data signaled an adverse process change (B89, B90). In other work, multiway PLS was utilized to analyze process variables and generate monitoring charts for on-line predictions of final product qualities (B91). Ku et al. developed a dynamic PCA method for detecting disturbances and isolating their sources in chemical processes (B92). Similarly, PCA as well as discriminant analysis were employed by Raich and C¸ inar for these purposes. Techniques for evaluating PCA model overlap and similarity, and the diagnosis of multiple disturbances were discussed (B93). Miscellaneous other statistical topics were discussed during the course of the last two years. For example, the semantics of true values and trueness of values were explained in the context of inorganic analysis (B94). The benefits of canonical correlation analysis over regression for the analysis of large amounts of data were also described by Chakravarty (B95). The analysis of variance and linear regression was utilized in assessing the Analytical Chemistry, Vol. 68, No. 12, June 15, 1996
homogeneity of silicon distribution in thin iron-silicon alloy strips (B96). Two experimental designs, a Plackett-Burman and a fractional factorial design, were applied to the HPLC determination of tetracycline hydrochloride from the USP XII. Several statistical interpretations of the designs for evaluating the assay were made and compared (B97). A number of toxicology applications of statistical methods were also reported in the literature. Erickson and McDonald argued that classical hypothesis testing may be inadequate for determining bioequivalence of control and test media in toxicological studies. As a result, they suggested alternative tests of bioequivalence (B98). Another study compared three statistical models for evaluating whole effluent toxicity data for regulatory compliance. They concluded that a mixed twoway ANOVA design with test crossed with concentration was best, and they provided a statistical test called the reliable toxicity detection level based on that model (B99). Vandeginste and Quadt described control charts for statistical quality control and emphasized their utility in facilitating communication between management and technical personnel (B100). A committee report on strategies for effective internal quality control was also given (B101). Outlier detection in multivariate analytical data was the subject of several papers. A genetic algorithm was shown to be a useful tool in the identification of outliers (B102). Alternatively, Walczak and Massart described robust principal component regression based on ellipsoidal multivariate trimming and least median of squares for detecting outliers (B103). In addition, Walczak proposed a new technique for outlier detection in multivariate calibration models using a subset selection scheme (B104, B105). OPTIMIZATION Optimization involves the minimization or maximization of a function of one or more independent variables. In so doing, one wishes to determine the values of these variables at the function optimum. Many methods have been devised to achieve this purpose. Among those methods most commonly utilized by chemists are the techniques of experimental design, simplex optimization, simulated annealing, genetic algorithms, and various classical gradient methods. For the chemist, the goal of using such methods may include the adjustment of system parameters for optimal method performance, wavelength selection for calibration or pattern recognition, or chemical structure estimation. This section delineates some of the current research that has been carried out by the science community in the development or application of optimization techniques. Several reviews of optimization were published in the course of this reporting period. Atkinson discussed recent developments in experimental design including topics such as restricted design regions, designs with both continuous and discrete factors, nonlinear models, multiple-objective designs, and more (C1). A review specifically of the Taguchi experimental design was given and its robustness toward environmental effects was explained (C2). Rozycki provided a review with an extensive list of references for the simplex technique in the context of the optimization of analytical methods (C3). Simulated annealing and its applications in chemistry were the subject of a review by Curtis (C4). In a brief overview of quality control concerns in the pharmaceutical industry, De Boer discussed several optimization approaches (C5). Ion-pair and ion-interaction chromatography were the focus of another review in which a number of references 28R
Analytical Chemistry, Vol. 68, No. 12, June 15, 1996
to optimization strategies were given (C6). Hayashi and Matsuda described the uncertainty structure of quantitative analysis and reviewed the use of information theory in the optimization of separation systems (C7). Information theory was also the subject of another review, with a particular focus on the use of the maximum entropy method in the analysis of time-resolved spectroscopy (C8). Finally, Lucasius et al. published part two of their comprehensive review of genetic algorithms (C9). As in previous review periods, a considerable number of the papers concerning optimization involved methods of experimental design. In particular, many of these methods were applied to the optimization of chromatographic systems. For example, de Vanssay et al. employed a Doehlert experimental design to optimize the gas chromatography (GC) conditions for the determination of permanent gases (C10). A Doehlert design was also applied to the optimization of variables in solid-phase spectrophotometry (C11). Mixture experimental designs were also utilized. They served to optimize the liquid/liquid extraction of drugs (C12) and to improve the sensory quality testing of mixed olive oils (C13). The estimation of a nonlinear calibration curve was achieved on the basis of a D-optimal experimental design (C14). In a two-part paper, a three-step procedure was reported for the optimization of liquid chromatography with respect to solvent system composition. This procedure combined the use of PLS regression and factorial design (C15, C16). Another chromatography application of experimental design was described by Guillaume and Guinchard. The optimization of the column efficiency in reversed-phase HPLC was achieved using a modified simplex method which reduced the number of required experiments by more than half relative to the general simplex (C17). The separation of alkaloids by overpressured layer chromatography was optimized on the basis of a PRISMA model in combination with a factorial design using the geometric mean of the normalized resolution as the objective function (C18). A number of researchers applied methods of experimental design in the area of toxicology as well. Michaud and co-workers described three types of experimental design and their potential pitfalls in the study of biological responses to chemical mixtures (C19). They also reported experimental design and data analysis methods for the investigation of toxic interactions in vitro (C20). In another study, experimental design was used to select optimum conditions such that the influence of experimental noise on the parameter estimates of toxicological models was minimized (C21). Hazardous waste was the focus of other work in which the effects of different sample pretreatment procedures on the determination of mercury in contaminated soil samples were investigated. A hierarchical experimental design led to the conclusion that sample pretreatment was a significant factor in the chemical analysis (C22). In a general article, the implementation of central composite designs was described and illustrated by several examples (C23). Langsrud et al. described a procedure for identifying significant factors in fractional factorial designs. They compared their technique with other established ones using a cheese data set (C24). A few workers utilized a Plackett-Burman experimental design for the optimization of instrumental analysis. Rogan et al. studied the effects of several system variables on resolution, retention time, and peak efficiency in chiral separation by capillary electrophoresis (C25). Similarly, lipid class analysis by HPLC was optimized with the help of a Plackett-Burman design (C26). In another paper, ruggedness tests of an HPLC
assay using two-level Plackett-Burman designs were reported. Various approaches to estimating the experimental error were presented (C27). The screening design of Taguchi was also employed by several researchers. This approach was used in optimizing analytical procedures in the analysis of cholesterol in foods, magnesium in feed premix, and moisture in mayonnaise (C28). The separation and detection of polychlorinated biphenyls (PCBs) by GC was also optimized using a Taguchi design in which the effects of several temperature and pressure parameters were investigated (C29). In addition, the photooxidation of dissolved organic matter in water samples was optimized by a Taguchi design for the analysis of trace metals by anodic stripping voltammetry (C30). In a series of papers, Lan and co-workers explained the theory of orthogonal array design for the optimization of analytical methods. Applications were given in atomic absorption spectrometry (AAS) (C31-C33), differential pulse polarography (C34, C35), and HPLC (C36, C37). Other applications of orthogonal array design included the optimization of microwave digestion for trace metal analysis (C38) and the minimization of the adverse influence of water on GC analysis (C39). Another chemometric method which has proved useful to a significant number of chemists is that of simplex optimization. Many applications and a few modifications of the simplex approach were reported in this review period. For example, several researchers applied the simplex method to the optimization of an inductively coupled plasma (ICP). Galley and others automated the optimization of four ICP parameters on the basis of the signal and signal-to-noise ratio (C40). In another study, spectral interferences in trace element analysis by ICP mass spectrometry were minimized using a simplex optimization with several response factors (C41). Similarly, the plasma parameters as well as the ion optics of an ICP mass spectrometer were subjected to a simplex analysis using the signal-to-noise ratio of many elements as the response function (C42). The electrothermal atomic absorption analysis of arsenic in environmental samples was optimized by considering several temperature, concentration, and time variables (C43). Other applications of the simplex method included the optimization of electrodeposition in R spectrometry (C44) and the enhancement of the preconcentration step in adsorptive stripping voltammetry (C45). In the latter study, a comparison with classical methods demonstrated that the simplex was more effective at optimizing the accumulation conditions. Several studies of the simplex methodology were also reported. A modified simplex was described for the optimization of response surfaces in experimental design (C46). In addition, a graphical procedure based on PCA was proposed for illustrating the progress of a simplex optimization. Several sample applications drawn from the literature were given (C47). A new simplex technique was reported for high-dimensional optimization. The method was demonstrated for peak resolution using synthetic Gaussian curves and actual infrared spectra (C48). In a novel approach, Andersson and Haemaelaeinen developed a method based on the simplex to adjust the retention times of chromatographic profiles. This was accomplished by optimizing the crosscorrelation between target peaks (C49). Optimization by simulated annealing and the Monte Carlo method proved to be quite popular among scientists. Woodruff and Bois compared the performances of Monte Carlo and simplex optimization for physiological toxicokinetic modeling (C50). In
addition, several researchers reported the use of Monte Carlo methods for the optimization of measurement systems. For example, Utteridge et al. simulated neutron depth dose distributions to improve prompt γ in vivo neutron analysis (C51). Likewise, the optimum backscatter geometry for measuring mercury and other heavy metals in vivo by X-ray fluorescence (XRF) was optimized using a Monte Carlo scheme (C52). Optimization of an annular-source excited XRF setup was also described (C53). In an application of simulated annealing, Beran and Szoeke showed that phase refinement in crystallography could be achieved with spatial constraints, unlike the Fourier methods, and do so with significantly less structural information (C54). A Monte Carlo method for crystal structure determination from powder diffraction data was also reported. The random displacement of atoms within a unit cell generated a series of structures which were evaluated on the basis of a Metropolis distance (C55). In surface crystallography by low-energy electron diffraction (LEED), Rous applied simulated annealing to the determination of surface structures. He also proposed a hybrid approach in which the simulated annealing performed the coarse global search and a conventional gradient method carried out the local refinement (C56). DNA solution structures were also calculated using a Metropolis Monte Carlo scheme from interproton and twodimensional NOE distance restraints (C57). In a nontraditional approach, protein three-dimensional structures and sequencespecific assignments of NOESY spectra were determined using simulated annealing without J-correlated spectra (C58). Backbone NMR resonances of labeled proteins were also assigned using simulated annealing (C59). In an environmental application, the problem of sampling spatially distributed contaminant levels was addressed using a multiobjective simulated annealing strategy (C60). A novel approach to the coating design for surface acoustic wave (SAW) sensor arrays was reported. Potential coatings were evaluated using a combination of a method called extended disjoint principal component regression (EDPCR) and Monte Carlo simulations of the SAW responses (C61). Ho¨rchner and Kalivas explained the theory of simulated annealing and generalized the step decision rule. In addition, they tested several variants of the algorithm using two global selectivity criteria in the wavelength selection of pure UV visible spectra (C62). Another group described a variable step-size generalized simulated annealing (VGSA) technique and applied it to the determination of parameters in an ion-selective electrode array analysis (C63). The optimization step in constrained background bilinearization was carried out by generalized simulated annealing (GSA). The performance of the latter was compared with that of conjugate gradient using both simulated and actual fluorescence data (C64). GSA was also used as the optimization method in a procedure for performing robust PCA using projection pursuit (C65). In an extension of the simulated annealing algorithm, the concept of messy chromosomes was introduced. Messy simulated annealing was shown to be a robust and efficient method for optimization of deceptive or multimodal functions (C66). Genetic algorithms constitute an important class of optimization methods inspired by Darwinian principles of evolution. Like simulated annealing, optimization is effected by a guided random search which renders them less prone to entrapment in local minima than the classical techniques. Moreover, genetic algorithms are characterized by an implicit parallelism which makes them particularly attractive for solving large-scale problems. Analytical Chemistry, Vol. 68, No. 12, June 15, 1996
Although initially a topic of considerable interest, fewer papers concerning optimization by genetic algorithms were reported in the last two years. Cong and Li explained the theory of a numeric genetic algorithm (NGA). They presented two architectures of their NGA and a new memory operator designed to accelerate convergence (C67). In a novel application, Martin et al. demonstrated a genetic algorithm for optimizing the design of an optical multilayer system (C68). A genetic algorithm was also used to search for a minimum-energy protein core sequence and structure from a rotamer library of potential hydrophobic residues (C69). Lucasius and co-workers described their genetic algorithm for the k-medoid clustering of large data sets and compared its performance with another technique known as CLARA on simulated data (C70). Two groups reported the use of genetic algorithms for the selection of wavelengths in multicomponent analysis. The first demonstrated that MLR combined with a genetic algorithm performed similarly to PLS regression in multivariate calibration (C71). The second compared simulated annealing, stepwise elimination, and genetic algorithms using Lorber’s selectivity and accuracy criterion, as well as Sasaki’s minimal squared error, as objective functions. Simulated annealing demonstrated the worst performance (C72). Ho¨rchner and Kalivas then showed that simulated annealing could indeed determine an optimum solution to this same problem provided wavelength searching was done in a close neighborhood of the previous wavelength subset (C73). Miscellaneous other optimization methods were also reported in the literature. Procrustes rotation and statistical variable selection were used for reducing the variable set in an industrial quality control application (C74). Subset selection was the topic of another paper in which a technique for optimum wavelength determination in multicomponent spectrophotometry was described (C75). Two groups reported the use of classical gradientbased methods. Powell’s conjugate gradient technique was utilized in conjunction with the mathematical model of a flow system to optimize flow injection analysis (C76). A second group coupled the steepest ascent method with multicriteria decision making to create a multiresponse steepest ascent strategy. The optimum was found by moving in the direction of a composite response made up of the individual response directions (C77). In addition, they described a steepest ascent strategy for multiple responses in a design space comprising both mixture and process variables (C78) Buice et al. introduced a nonparametric method called universal numeric optimization (UMO) for optimizing the instrumental parameters of an acoustic resonance spectrometer (C79). Biochemical structure determination was also the focus of optimization methods. Olson and Markley devised an algorithm for the sequential assignment of NMR backbone resonances of proteins (C80). Another group presented a variable-target intensityrestrained global optimization (VARTIGO) method for estimating three-dimensional polypeptide structures from NOESY data. Application of this method to a model peptide resulted in excellent agreement with experimental data and indicated that it may be superior to the intensity-restrained single-target function (STF) technique (C81). Application-specific methods were also described for the optimization of transient signal measurements in ICPMS (C82), as well as the optimization of mixture composition for IPG casting with a two-chamber mixer (C83). Several reports on the use of mathematical programming for optimization of chemical analysis were also found. Westrich et al. described linear and quadratic programming for estimating nutrient values in food 30R
Analytical Chemistry, Vol. 68, No. 12, June 15, 1996
products (C84). Hudak generated monitoring configurations for groundwater contaminant detection by using a procedure that combined simulation and linear programming (C85). Similarly, a simulation model coupled to nonlinear programming was used for the optimization of liquid chromatography (C86). Another technique known as statistical scanning was reported for optimal mobile-phase selection in HPTLC (C87-C90) and HPLC (C91). Two other popular optimization methods in chromatography, those of overlapping resolution maps (ORM) and window diagrams, were also reported. The ORM method was used to select the optimum mobile-phase composition in HPLC (C92), while window diagrams were employed for multicolumn optimization in high-speed GC separations (C93). A new method based on the Hooke-Jeeves algorithm, known as optimization procedure by search point (OPSP), was devised for finding optimal HPLC conditions. The OPSP outperformed the modified simplex and the grid search procedures in terms of the number of experiments required to determine the optimum solvent composition (C94). Target factor analysis was found to be useful for optimizing capillary electrophoresis conditions with only a small number of experiments (C95). An experimental procedure may consist of two or more steps, each of which needs to be considered for the optimization of the overall process. Lundstedt and Thelin presented a scheme consisting of experimental design, PCA, multivariate design, and PLS regression for the optimization of a twostep procedure (C96). Kettaneh-Wold and co-workers proposed modifications to PLS for the multivariate design of process experiments (M-DOPE). Past and current information were combined to form orthogonal groups of variables for experimental design and optimization (C97). Information theory also played a role in the optimization of chemical analysis. Shannon’s entropy method was explained and illustrated with some applications to problems in analytical chemistry (C98). In addition, Frederick described the use of maximum entropy in determining the vibrational frequencies of chemisorbed species (C99). A numerical technique called conditional maximization (CM) was explained for calculating maximum likelihood estimates in the absence of closed form expressions. Further, the utility of combining CM with the expectation/minimization (EM) method was also shown for dealing with incomplete data (C100). In a new approach, Power and Prystay developed a solution to the ill-conditioned problems in photothermal science. This method was based on the expectation minimum principle, which adds known random noise to a model basis in order to enhance its conditioning. Fitting to the experimental data was accomplished by linear minimization or projection schemes (C101). In a novel enhancement of an optimization method, feed-forward neural network weights were trained by an adaptive, global, extended Kalman filter. This methodology achieved significantly better convergence properties than its nonadaptive counterpart by circumventing calculations with data containing redundant information (C102). SIGNAL PROCESSING This section is concerned with research in signal processing, that is, with the development or use of methods for enhancing analytical measurements with respect to chemically or physically relevant information. In the course of the two years covered by this review, work in this area has mainly involved time series analysis, digital filtering, smoothing, deconvolution, background correction, and image analysis. Numerous methods were inves-
tigated in this regard. These included correlation techniques, various filters, and transform methods, as well as information theory. Several reviews of signal-processing methods were published during this reporting period. For instance, Sande provided an overview of time series analysis of optical and crystallographic data and discussed the problem of phase retrieval. Methods included iteratively reweighted least squares and convolution for accessing the Fourier transform and radial projection (D1). In another review, Grunwald and Herzog described the use of the Fourier transform for explaining dynamic spectroscopy and provided a fast Fourier transform algorithm (D2). Both Fourier and Hadamard transforms were also explained in the context of spectroscopy by Graff (D3). The benefits of combining difference measurements and shaped waveforms with Hadamard transform FT-ICR mass spectrometry were described by Ha¨bel and Ga¨umann (D4). Another review was provided on the issue of spectral interferences in ICP-AES. Topics included line selection, detection limits, and background correction (D5). Signal-processing reviews were also published with respect to NMR spectrometry. Signal averaging and digital filtering in FT NMR were explained by Fuson (D6), while Borer and co-workers reviewed the maximum likelihood method for deconvolution of multidimensional NMR spectra (D7, D8). The correlation analysis of time series was applied to a number of different measurement systems. Two studies investigated the dynamics of polymer-coated gas sensors. In one, the parameters of the system were estimated by using an autoregressive model whose coefficients characterized the target gas. Both pure and mixed gases were identified and quantified by this method (D9). In another gas sensor study, the Volterra and Wiener theories were briefly discussed in order to explain the use of kernel functions for modeling dynamic systems. Cross-correlation methods in nonlinear modeling were described (D10). Zhang and Phillips reported a multiplex GC technique for the determination of organic compounds in solids. Multiple signals from a thermal desorption modulator at the head of the capillary column led to a multiplex detector signal from which the chromatograms were retrieved by cross-correlation calculations (D11). Aroclor and Aroclor mixture chromatograms from GC/MS were also subjected to time series analysis. Fitting the experimental autocorrelation function with theoretical models permitted the determination of the number of PCB congeners present in a specific Aroclor (D12). In another report, a Fourier transform-based cross-correlation method for the automatic detection of spectral information in microliter volumes was described. The method involved the crosscorrelation of interference-free and noise-free masks with spectral patterns from multielement mixtures (D13). Two other studies utilized the cross-correlation of experimental and library data for the interpretation of spectra. Eng and co-workers devised a strategy for identifying peptides by correlating their tandem mass spectra with fragment ions predicted from amino acid sequences in a database (D14). Doerffel and Kreher compared the shape of the cross-covariance function between the infrared library and sample spectra with that of the autocovariance function of library spectra. In this manner, analytes were identified even at low signal-to-noise ratios (D15). Doerffel also described the detection of long-term trends in noisy bivariate data by a two-dimensional cumulative sum technique called 2D-CUSUM (D16).
A significant number of the papers in the area of signal processing concerned the enhancement of analytical signals using smoothing and digital filtering techniques. Various smoothing and differentiating Savitsky-Golay filters were investigated and compared in terms of the total error criterion. The benefits of multipassing quartic/quintic polynomials for smoothing and cubic/quartic polynomials for differentiation were demonstrated. In addition, a means of selecting the best polynomial length in the fine-structure analysis of composite spectra was given (D17). An adaptive-degree polynomial filter was proposed to circumvent the need for the a priori selection of the polynomial degree in Savitsky-Golay filtering. This was accomplished by incorporating a heuristically guided statistical test for the polynomial degree in each window. The adaptive method was superior to that of Savitsky-Golay on synthetic noisy data (D18). Kim and Marshall described a method for resolution enhancement of Fourier transform spectra, without loss of signal-to-noise ratio, using magnitude-mode multiple-derivative spectra. Their approach was based on a time domain exponential window (D19). Several other studies focused on the enhancement of infrared spectra. A modified least-squares technique was presented for the smoothing and differentiation of FT-IR spectra of coal (D20). Another report concerned the use of second-derivative spectra in self-modeling mixture analysis. Ways of obtaining optimal smoothing of NearIR and FT-IR microscopy data were given (D21). Riris and others discussed the enhancement of data from near- and mid-infrared diode laser sensors using several signal-processing methods such as digital bandpass filters, Wiener and matched filters, and leastsquares fits (D22, D23). In a Raman spectroscopy application, an interactive polynomial filter with variable bandpass was devised using a least-squares approach. This filter was demonstrated in the detection of weak peaks in a noisy spectrum, and the detection of minor component signals superimposed on major ones (D24). Fourier transformation of multiple injection signals from FIA measurements were discussed for improving the signal-to-noise ratio (D25). Lee and co-workers generalized the method of Fourier smoothing of FIA data by supplementing the standard complex exponential basis functions with other complete orthogonal sets such as the Gram or Meixner polynomials. In addition, a generalized Akaike information theoretic criterion was utilized to determine the best filter order (D26). A new filter based on the Fourier transform of the second derivative of a Gaussian function was proposed for the enhancement of overlapped spectral lines (D27). Work was also performed on the improvement of electrochemical data. Legendre polynomials were employed for data reduction and noise filtering of amperometric signals (D28). Four signal-processing procedures, namely, moving-average smoothing, polynomial smoothing, rectangular low-pass filtering, and exponential low-pass filtering were compared for use in potentiometric stripping analysis. The rectangular low-pass filtering technique was most effective in enhancing the resolution of overlapping peaks (D29). Stripping voltammetry data were also subjected to signal processing such as background subtraction, ensemble averaging, digital filtering in the time and frequency domains, multiple scanning, and deconvolution (D30). In addition, finite impulse response (FIR) and infinite impulse response (IIR) filters were employed for signal-to-noise ratio enhancement (D31). Noisy Auger spectra were filtered using two new shapes for transmission functions in Fourier space. The performance of these functions was reported to be superior to that of an optimal Analytical Chemistry, Vol. 68, No. 12, June 15, 1996
transmission function obtained from the literature (D32). In another study, a procedure based on linear digital filtering and nonlinear least-squares fitting was described for quantitative Auger analysis of pseudobinary compounds (D33). Enhancements made to a particle-induced X-ray emission (PIXE) software package called GUPIX were reported. These included modifications to the nonlinear least-squares fitting procedure, systematic weighting for dealing with uncertainties in low-energy tailing, and recommendations for choosing appropriate filter dimensions (D34). In a separations application, ensemble averaging and digital filtering methods were investigated for signal-to-noise ratio enhancement of data from chromatography and electrophoresis (D35). A comparison was made between signal-processing procedures and those typically employed for atom probe data noise reduction. The former included a median filter and a double-window modified trimmed mean filter (D36). Finally, a moving median filter was applied to mass spectral and potentiometric data. The filter removed outliers without significant distortion of the signal while enhancing the signal-to-noise ratio (D37). The Kalman filter continued to find favor with a number of workers as a tool for signal processing. Riris et al. utilized the Kalman filter for decreasing the relative noise of tunable diode laser absorbance measurements by 1 order of magnitude (D38). In an ICP-AES application, the Kalman filter was employed to model the emission and remove the background using a library of pure-component scans (D39). Deng and co-workers used a fast Fourier transform/stochastic model of the log-fluctuating conductivity from tracer data at a groundwater site. A Kalman filter adaptively provided parameter estimates as the tracer evolved (D40). Signal smoothing by cubic spline functions was optimized using a Kalman filter to select the best knot distribution and the values of the power basis coefficients (D41). In another study, an information theory approach for the optimization of an adaptive Kalman filter was described by Agbodjan and Rutan (D42). A considerable number of papers regarding deconvolution methods appeared during the course of this review period. Ferry and Jacobsson proposed a simulated annealing approach in which information about the instrumental broadening was incorporated in the cost function to be minimized. Thus, effects of the finite instrumental resolution were eliminated from the fit of Raman spectra (D43). Shifted-spectra, edge detection, and fast Fourier transform filtering and deconvolution methods were also utilized for the rejection of fluorescence and the improvement of resolution in Raman spectra (D44). In another report, the enhancement of IR spectra by Fourier deconvolution using Cauchy-Gauss product functions was described. Two procedures for determining the inverse Fourier transform, a numerical and an analytical one, were given (D45). The resolution enhancement of IR spectra was also discussed by Mantsch and Moffatt using Fourier self-deconvolution and Fourier derivation (D46). In a chromatography study, the sharpening of peaks by reducing the contribution of a blurring function to the signal was investigated using a filter for deconvolution (D47). Fourier deconvolution was also employed for processing liquid chromatograms coupled to flow radioactivity counters (D48). In another chromatography application, the use of exponentially modified Gaussian functions for modeling and deconvolving overlapped asymmetric peaks was described (D49). Several workers reported deconvolution methods for NMR data as well. Maudsley discussed spectral line shape estimation by self-deconvolution . The method accurately determined amplitude 32R
Analytical Chemistry, Vol. 68, No. 12, June 15, 1996
and phase line shape distortions possibly caused by field inhomogeneity and gradient eddy-current effects (D50). In addition, fast Fourier transform dePaking was described for line shape determination (D51). Deconvolution was also utilized in a hybrid/ hybrid matrix scheme for three-dimensional NOE/NOE data (D52). In electrochemistry, analysis by adsorptive stripping voltammetry was reportedly improved by means of signal deconvolution (D53). In mass spectrometry, a patent application was submitted for a method designed to eliminate noise and artifacts in the spectral deconvolution (D54). A multiple convolution/ deconvolution scheme was developed to remove broadening factors caused by the secondary ion mass spectrometer and the sample itself in the analysis of solid-source diffusion from disilicide sources (D55). The deconvolution of molecular beam time-offlight waveforms using Fourier transform and Weiner filters was the subject of another study. No restrictions were imposed on the functional form of any factors influencing the shape of the time-of-flight distributions (D56). Low-count nuclear spectra were deconvolved using a probabilistic approach. The method was utilized for computing small peak areas on high background, and for resolving overlapping peaks (D57). A new technique described as digital moving window deconvolution was reported for processing signals in high-resolution, high-throughput γ-ray spectroscopy (D58). Another deconvolution-based procedure was developed for enhancing XPS spectra by removal of instrumental artifacts (D59). Busse and co-workers described procedures for deconvolving electron spectra and warned against the indiscriminate use of methods such as fast Fourier transforms with Wiener filtering (D60). Backscattered electron signal profiles of bulk specimens by SEM are prone to distortion due to the beam size, the detector, and the sample itself. A deconvolution technique to enhance the spatial resolution was given (D61). Similarly, the lateral resolution of electron probe microanalysis was improved using special digital filters (D62), and a Fourier deconvolution (D63). Several papers were published describing the use of MLE for the deconvolution of spectra. Liew and co-workers reported a MLE procedure for recovering trace element depth profiles from PIXE data. The method was based on the exact stochastic model of X-ray yield measurements (D64). Borer and Levy enhanced the resolution of three- and four-dimensional NMR spectra and reduced the noise and cross-talk between planes using an MLEbased deconvolution (D65). Digital filtering and MLE were also applied to 13C NMR spectra of multicomponent mixtures (D66). In electron energy-loss spectroscopy, Frederick et al. compared several deconvolution methods. These included a direct technique involving a linearized maximum entropy filter and iterative techniques based on maximum likelihood, maximum entropy, and Bayesian principles (D67). Another comparative study of deconvolution techniques was carried out in which six of the most common procedures were evaluated using synthetic calorimetric data, as well as synthetic and real spectrometric data (D68). The issue of background correction was addressed by a significant number of researchers. Dobrowolski and Jamroz presented the intensian method, a generalization of the BeerLambert law, for the elimination of background in 3-dimensional spectroscopy (D69). Several background correction techniques including linear, Shirley’s, Tougaard’s, and deconvolution methods were explored for XPS and were found to lead to similar results (D70). Another comparative study of background correction procedures was made in XRF spectroscopy. Background models
included exponential functions and linear combinations of polynomials, while estimation strategies involved an iterative stripping algorithm and a channel selection procedure (D71). An iterative approach based on splines and smoothing was also employed for XAFS data (D72). In response to a proposed robust estimation method for the elimination of spike noise in charge-coupled device data, Hill argued that missing-point fitting is far superior (D73). Baseline correction in NMR spectroscopy was investigated by several workers. Brown provided an automatic procedure for the correction of baselines in one- and two-dimensional NMR spectra using Bernstein polynomials (D74), while Levy et al. automated the correction of base-plane and systematic noise in twodimensional NMR spectra through a five-step procedure (D75). An iterative algorithm for the simultaneous base point correction and signal recognition in multidimensional NMR was also reported (D76). Kaljurand et al. discussed the correction of baseline drift and noise removal using a piecewise linear function and low-pass digital filtering in correlation chromatography (D77). Chromatographic baseline noise from a variety of detectors was also investigated in relation to the determination of detection limits. Signal-processing methods included FIR filters and nonweighted moving-average smoothers (D78). Johnston studied the effect of detector baseline offsets on the determination of fluorescence modulation lifetimes. He provided a plotting technique for both detecting baseline error and computing accurate fluorescence lifetimes in the presence of such error (D79). The theory of repetitive wavelength scanning and Fourier analysis in AES for background correction was described by Matthee and Visser (D80). Further, Coles et al. introduced a correlated background correction (CBC) procedure to predict the background emission under transient analyte peaks in ICP-AES (D81). Two procedures for automatic background estimation, one heuristic and one statistical, were also described for ICP-AES (D82). In ICPMS, a method for interferent correction based on some a priori knowledge of the sample matrix was presented (D83), as well as a procedure for mass bias drift correction using an internal standard (D84). The background noise correction of line spectra was discussed by Iwata and Koshoubu. They developed a numerical procedure based on singular value decomposition (SVD) for extracting the true signal without distortion of the waveform (D85). A combination of interferogram segment selection, digital filtering, and pattern recognition was utilized for suppression of the variable detector envelope in remote IR detection of volatile organic compounds (D86, D87). Bandpass digital filters were also employed to extract analyte-specific information from short FTIR interferogram segments in another study. Univariate calibration models were generated without separate background or reference interferograms (D88). A procedure for optimizing the design of the bandpass filters was also described (D89, D90). An artificial neural network with optimal associative memory (OAM) was reported for background correction of single-scan IR spectra. Both univariate and multivariate calibrations were improved relative to the uncorrected spectra (D91). Bidirectional associative memory (BAM) and modified BAM were compared to OAM for background correction performance in another study (D92). Wang and co-workers explained the influence of stray light on the spectral background and its detrimental effect on multivariate standardization. Simulated and real data were employed to describe the correction of the additive background term (D93).
Fourier transform methods were also utilized in a number of other applications. Rockwood explained that molecular isotope distributions can be obtained from the multiple convolution of elemental isotope distributions. He introduced a Fourier transform approach to the calculation of these distributions which was more efficient than the usual polynomial expansion method (D94, D95). A nonuniform grid Fourier transform analysis, coupled with filtering and geometric localization techniques, was applied to the correction of electron beam proximity effects (D96). In another report, the analysis of complex nuclear spectra was performed based on power spectra estimated from modified discrete Fourier transforms. The resolution enhancement achieved was comparable to that of a common peak-fitting procedure called SAMPO (D97). A method for spectral improvement by Fourier thresholding (SIFT) was described for reducing the relative noise in time-resolved spectra (D98). In an NMR application, a comparison was made between time and frequency domain analyses for parameter estimation. A combined time and frequency domain processing strategy was proposed (D99). The coefficients from a generalized Fourier expansion into a weighted linear combination of discrete orthogonal polynomials were used to represent digitized transient signals. Both Gram and Laguerre polynomials were studied, and the utility of the method was demonstrated on real and simulated FIA data (D100). Although Fourier transformation is by far the most common, other transforms were also employed for processing analytical signals. Procedures for generating equivalence classes of Hadamard matrices were given (D101). A Hadamard spectroscopic imaging method insensitive to pulse imperfections was also presented (D102). Three techniques, namely, direct, Hadamard, and Fourier were described for ESR imaging based on the modulation-field phase (D103). Another type of transformation called the Kramers-Kronig transform was also used in chemical analysis. Diaz et al. applied the Kramers-Kronig transform to electrochemical impedance data for characterizing damaged automotive paint (D104). Kramers-Kronig transforms were also utilized in IR reflection spectroscopy for determining the complex refractive index of anisotropic materials (D105). Wavelet transforms have recently generated significant interest in the statistical and engineering communities and have begun to appear in the chemistry literature. IR spectra of substituted benzenes were subjected to a transformation based on Daubechies wavelets for the purpose of feature extraction. The substantial reduction in data resulted in improved classification performance (D106). Similarly, wavelet transforms were applied to FT-IR spectra for estimating the mass fraction of minerals in rock (D107). Finally, a method based on the wavelet transformation was described for determining relative ion abundances in ICR-MS (D108). The maximum entropy method (MEM) has proven to be a useful tool for a number of researchers in analytical chemistry. The attractiveness of MEM lies in its ability to perform signalprocessing tasks such as line sharpening, noise suppression, and deconvolution without the assumption of an underlying model. Maximum entropy data analysis was explained by von der Linden (D109). McGown and co-workers applied MEM to total lifetime distribution analysis in fluorescence spectroscopy. They discussed the advantage of MEM over nonlinear least squares with respect to model bias (D110). The deconvolution of dielectric and impedance data using MEM was also described. The authors of this paper, however, concluded that in most cases nonlinear least Analytical Chemistry, Vol. 68, No. 12, June 15, 1996
squares was preferable to MEM with this type of data (D111). In a mass spectrometry application, MEM was utilized for the deconvolution of secondary ion mass spectra. A 10-fold increase in resolution was achieved without a concomitant loss of signal intensity (D112). The resolution of simulated and real electron energy-loss spectra was enhanced by several methods for comparison. These included MEM and a variety of filtering techniques (D113). The deconvolution of XPS spectra using MEM was described, and the advantages of this approach over the more common Fourier transform method were discussed (D114). Angle-dependent XPS data were simulated in order to develop an MEM-based method for the determination of depth profiles. Overall scaling of the noise, confidence limits, and relative probabilities of the models were calculated (D115). In addition, angle-resolved XPS data were subjected to MEM for the reconstruction of composition depth profiles while accounting for the effects of elastic scattering (D116). In a Raman spectroscopy application, two-point maximum entropy was employed for the signal-to-noise ratio enhancement and deconvolution of the spectra. This strategy, which required no filter parameters or a priori knowledge of the signal, was compared to the SavitskyGolay method (D117). A polemic in response to a paper entitled “What is Wrong with MEM?” was put forward in which the authors systematically refuted the arguments against MEM for spectral estimation (D118). Several reports regarding image analysis techniques appeared during this review period. A deconvolution contrast procedure for the resolution of chemical species in NMR images was described, and its advantages over other approaches were explained (D119). X-ray and γ-ray computed tomography (CT) images were reconstructed using the summation convolution backprojection (SCBP) technique. This method was compared to direct Fourier reconstruction (D120). Boehmig et al. introduced a new scheme based on the k-means algorithm for matching scanning Auger microscopy (SAM) images which were measured at different times or positions of the sample (D121). In a study of the pore structure of activated carbon fibers, transmission electron microscopy (TEM) images were enhanced using noise reduction, low-frequency cutoff filtering, and binary image formation. Two-dimensional Fourier transformation was utilized for analysis of the porosity size distribution (D122). Several extensions to multivariate image analysis (MIA) were introduced including data preprocessing, synthetic multivariate image models, and visualization tools. The extended MIA technique was demonstrated using simulated and real positron emission tomography (PET) data (D123). Lindgren and co-workers extended their kernel-based PLS method to include cross-validation and applied it to the analysis of multivariate images (D124). Miscellaneous other signal-processing techniques were also described in the literature. Multiplicative signal correction was applied to Raman spectra to improve the calibration of an anesthetic sensor by removing multiplicative and additive effects from the data (D125). Scale differences were demonstrated in near-IR reflectance spectra depending on the order of the application of standard normal variate (SNV) and de-trend transformations. In addition, the relationship of multiplicative scatter correction (MSC) to the SNV transformation was given (D126). Helland and co-workers also discussed the connection between MSC and SNV transformations, as well as several variants of MSC. These preprocessing methods were compared using eight mul34R
Analytical Chemistry, Vol. 68, No. 12, June 15, 1996
ticomponent data sets analyzed by diffuse near-IR spectroscopy (D127). Penninckx and others described a scheme for the detection of interferences in AAS based on Hotelling’s T2 test statistic (D128). In a different approach, two nonparametric smoothing techniques for detecting interactions in high dimensional data were presented (D129). In another study, a new technique for the analysis of multicomponent exponential decays based on multiple differentiation was presented (D130). Similarly, a method was described for determining phosphorescence lifetime distributions in heterogeneous systems. The procedure utilized a quadratic programming approach to fit the data to linear combinations of exponentials (D131). Cook derived a relationship between the truncation length of an FT interferogram and the distortion of the peak half-width to determine the influence of path length on line shape (D132). In a general approach, inhomogeneous line shape determination in EPR spectroscopy was described using a convolution method based on the LevenbergMarquardt algorithm (D133). Flanigan examined the dependence of detection sensitivity on spectral resolution in the remote detection of organic vapors. In particular, the signal-to-noise ratio was calculated for an SF6 target band modeled as a Lorentzian (D134). Fourier transform ICR and NMR spectra were enhanced with respect to resolution and signal-to-noise ratio using a data reflection algorithm. The improvements were a consequence of phase matching (D135). Several reports were published on the use of artificial neural networks (ANN) in the processing of data from chemical sensors. A radial basis function network was developed for a SAW sensor array. Compensation for response time, a major concern with gas sensors, was addressed through a deconvolution algorithm (D136). Similarly, Endres et al. investigated the processing of sensor array signals with an ANN and discussed the problem of the time-dependent behavior of sensors (D137). In addition, they devised an empirical drift correction scheme and presented a new method called the dynamic test point distribution (DTPD) for improving the calibration (D138). In a novel approach to error prediction, Hayashi and Matsuda developed a theory to predict the precision of measurements in HPLC. A mixed stochastic process of white noise and Markov chains was utilized to model the baseline drift, and parameters were estimated by a least-squares fit of the power spectral density of the baseline (D139). Alsberg and co-workers discussed the compression of high-dimensional data using Bsplines for computational space and time savings. In particular, they applied the method to second-order FT-IR spectra to demonstrate that the PCA results of the compressed data were comparable to those of the original data (D140). Finally, although signal-processing methods were not the focus of the work, it should be noted that a large number of papers applied such methods to the structure determination of biological molecules. Only a few are mentioned here. For example, the secondary structure of proteins was estimated on the basis of Fourier self-deconvolution, differentiation, and curve fitting of FTIR spectra (D141). The effect of high pressure on the secondary structure of proteins was also investigated by FT-IR using deconvolution and fitting methods (D142). In another study, the three-dimensional structure of the hydra head activator neuropeptide was modeled with the help of a double-iterated Kalman filter (D143). Another paper on structure determination described a fast Fourier inversion procedure for the three-dimensional reconstruction of helical structures (D144).
RESOLUTION This section is concerned with chemometrics methods for resolution and recovery of pure-component spectra from the overlapped spectra of mixtures. Articles dealing with pure spectral deconvolution (i.e., elimination of the slit function effect) appear in the section on Signal Processing. During the last two years, Brereton published a review of mixture deconvolution by PCA (E1), and resolution methods for multicomponent kinetics have also been discussed (E2). Another review invoked the application of chemometrics approaches for source apportionment in environmental problems concerning polychlorinated dibenzodioxin and polychlorinated dibenzofuran (E3). First-order analytical data have mainly been resolved by PCAbased techniques. A group of geologists proposed a new technique, polytopic vector analysis, to establish the identity and relative contributions of polychlorinated dibenzo-p-dioxins and polychlorinated dibenzofurans in surface and near-surface sediment samples (E4). Jiang et al. presented a new two-step resolution technique, using the criteria of the best proposed solution and SVD to process an ill-conditioned matrix system (E5). SVD followed by iterative target transformation analysis (ITTFA) was used for the simultaneous determination of 15 rare earth elements from ICP-AES data (E6). The assessment of peak purity is often a concern in chromatography. A new method was presented using condition index evolving profiles and singular value evolving profiles (E7). On the same subject, Cuesta Sanchez et al. applied eigenstructure tracking analysis (ETA) and compared the results with those obtained by fixed size window evolving factor analysis with SIMPLISMA (another iterative curve resolution method), and with methods based on Gram-Schmidt orthogonalization (E8). Liang and Kvalheim applied ETA to unresolved chromatographic peaks, producing a rank map which is subsequently used to define a set of orthogonal projection matrices. This procedure enables unambiguous identification of the peak pattern in an unresolved region (E9). Gurden et al. suggested a new factor analysis (FA) technique using spherical projections for resolution and recovery of mid-IR mixture spectra as an alternative to the conventional eigenanalysis (E10), and they studied the influence of noise and other parameters on the method (E11). They also proposed a method based on the successive elimination of single-component spectra determined by movingwindow PCA (E12). A method called real factor analysis was introduced as a new self-modeling technique for the deconvolution of data from kinetics studies. The method, claimed to be an alternative to eigenanalysis, was successfully tested on simulated data matrices (E13). Derivative spectroscopy is now commonly used by the analyst to enhance resolution. Windig discussed the calculation of second-derivative spectra with respect to selfmodeling mixture analysis techniques and proposed methods to obtain optimal smoothing (E14). Resolution applied to higher-order data (second and third order), with a special emphasis on chromatographic techniques hyphenated to spectroscopic techniques, remains the main source of papers in this section. Tauler discussed the application of multivariate curve resolution to second-order hyphenated LC with different data structures depending on the reproducibility of the elution process (E15). Liang et al. proposed a new procedure for the separation of embedded peaks in multiwavelength chomatography of binary mixtures, based on a rank analysis of the firstderivative matrices (E16). Rank analysis of orthogonal projections
was also used for nonlinear curve fitting in GC/IR spectrometry and variable-temperature IR studies (E17). Phillips and Georghiou described trilinear curve resolution applied to polarized fluorescence spectra (E18). In thin-layer chromatography, overlapped amino acids were detected by applying the direct trilinear decomposition method (TLD) (E19). TLD also allowed resolution of second-order emission/excitation fluorescence spectra of animal dental calculus (E20). Elbergali et al. discussed the problem of wavelength selection and resolvability indexes for clusters of peaks in FA of HPLC/DAD data. Their approach was applied to mixtures of chlorophyll degradation products (E21). Neal proposed an FA-based algorithm using distance as a measure of spectral similarity applicable to data produced by hyphenated and multidimensional techniques (E22). An article by Tauler et al. presented an overview of multivariate curve resolution techniques and a new method for self-modeling curve resolution (SMCR) was described, this one based on a simultaneous analysis of different data matrices with one or two orders in common, using constrained alternating least-squares optimization (E23). Evolving factor analysis (EFA) is an iterative, PCA-based approach for initial estimation of concentration profiles used in one application for the determination of the chemical species formed in the Fujiwara reaction (E24). Another iterative procedure is parallel factor analysis (PARAFAC), a resolution algorithm based on alternating least-squares optimization that can be used on trilinear data arrays. In some instances, PARAFAC can become trapped in a “swamp”sa region of inferior resolution for some time before converging to an acceptable resolution. This behavior has been investigated and solutions have been proposed (E25). SIMPLISMA was used in synchronous fluorescence spectroscopy (E26) and for the assessment of peak purity in LC/DAD (E27). Rank annihilation factor analysis (RAFA) was successfully used to predict the analyte concentration in one- and two-component samples of hydroxybenzaldehydes analyzed by FIA/DAD. In both cases, the apparent rank of the data matrices did not correspond to the number of chemical components in the samples (E28). However, when the retention times exhibit some shifts, RAFA fails to produce the correct concentration estimates. Grung and Kvalheim proposed a correction method based on Bessel’s inequality to adjust for the retention time shifting of the standards (E29). The theory, properties, and implementation of the generalized rank annihilation method (GRAM), a curve resolution and calibration technique derived from RAFA, were discussed by Faber et al. in three consecutive papers (E30-E32). GRAM was also used to eliminate background signal and separate partially resolved peaks in HPLC/ DAD spectra of drugs of abuse in urine (E33). Window factor analysis (WFA) is a noniterative, PCA-based curve resolution method. The influence of the choice of the elution regions on the performance of WFA was investigated by Elbergali and Brereton (E34). The same authors carried out simulations to evaluate the influence of noise, peak position, and spectral similarities on resolvability of HPLC/DAD data by WFA (E35). Other simulations allowed testing of a method for selecting variables in WFA, using the calculation of resolvability indexes and a double-window technique (E36). Three methods for eigenanalysis in WFA of two-way data matrices were described and compared (E37). The performance of heuristic evolving latent projection (HELP) and EFA was assessed for the resolution of simultaneous dynamic thermal processes in thermochromatography of oil shale (E38). HELP was favorably compared to EFA Analytical Chemistry, Vol. 68, No. 12, June 15, 1996
and the method of orthogonal projections for the resolution of simulated multicomponent systems with partial selectivity (E39). Another simulation showed that HELP performed better than ITTFA and alternating regression. In this study, a multiarray resolution parameter was introduced to evaluate the performance of each method (E40). Neal explored the use of correspondence analysis, a variant of nonlinear principal component analysis, for the resolution of nonbilinear two-dimensional COSY NMR data (E41). Resolution of analytical signals can also be achieved by curvefitting overlapped peaks with a Kalman filter. Evolving projection analysis (EPA), formerly called evolving principal component innovation analysis, is a rank analysis method that can handle nonlinear responses and nonuniform measurement noise. A network of parallel Kalman filters can be used for recursively modeling absorbance in ordered data sets like spectrochromatograms (E42). It was applied by Wentzell and Hughes to secondorder data sets consisting of more than two components (E43). Adaptive Kalman filtering was used for resolving components from UV/visible spectra in solvatochromic studies (E44, E45) and room-temperature phosphorescence spectroscopy (E46). Accuracy and precision of Kalman filtering for the resolution of data from first- and second-order kinetics were also evaluated (E47). Another type of filtering has been applied to the quantification of overlapping chromatographic peaks. On the basis of an a priori knowledge of the shapes and positions of the peaks, the authors used a matched filter followed by a simplex optimization of the output parameters. The performance of the method was reported to decrease as the degree of overlap and the signal-to-noise ratio increased (E48). A few articles mentioning the use of other statistical methods for resolution appeared in the last two years. Reh compared the performance of nonlinear regression and iterative linear approximation for the peak-shape analysis of chromatographic data and made some suggestions for the use of these tools in practical applications (E49). Faber et al. discussed several pseudorank estimation methods that make use of prior knowledge of the size of the measurement errors and suggested a t-test on singular values to increase the significance level of the pseudorank estimation (E50). The same authors discussed the pseudorank estimation methods based on the eigenvalues obtained from PCA and used Malinowski’s reduced eigenvalues to construct an F-test for the determination of the number of primary principal components (E51). Another theoretical study involved the examination of the elements of a dispersion matrix obtained in a GaussNewton nonlinear least-squares procedure to evaluate the reliability of the parameters in the resolution of overlapped Gaussian peaks (E52). Several papers have also been recently published mentioning the use of optimization procedures for resolution. Bramanti et al. presented a new technique based on a constrained conjugate gradient minimization algorithm applied to second-derivative FTIR spectra (E53). In DSC, a simplex optimization of peak parameters allowed the separation of overlapped model peaks (E54). The flexibility and modeling power of neural computation are now commonly exploited in several areas of chemistry, and resolution is no exception. A Dutch group applied ANN to peak detection and curve fitting of X-ray diffraction spectra (E55). Another approach is based on the concept of sensor fusion data analyzed by ANN. The results obtained were as good or better 36R
Analytical Chemistry, Vol. 68, No. 12, June 15, 1996
than those obtained when a mathematical curve-fitting technique was used (E56). Nonlocal curve fitting of infrared spectra containing strongly overlapping peaks was performed by a genetic algorithm-based procedure (E57). Finally, the Gram-Schmidt orthogonalization procedure allowed the segmentation of features of interest in magnetic resonance imaging (E58). An algorithm based on this technique was described for the assessment of peak purity from LC/DAD data (E59), and modifications of this algorithm were proposed in a subsequent paper (E60). CALIBRATION Calibration involves relating, correlating, or modeling a measured response based on the amounts, concentrations, or other physical or chemical properties of a set of analytes. Multivariate calibration, the measurement of analyte concentrations in mixtures of several components, remains by far the fastest growing area of chemometrics. A tremendous number of papers have appeared in the last two years on the topic of calibration. In almost all of these applications of multivariate calibration, the authors employed now well-established methods such as PLS regression and PCR. The great diversity in the application of these calibration methods indicates that these methods are being increasingly adopted by research groups far removed from the field of chemometrics, a consequence of the interest that chemometrics is attracting from far-removed fields and of the increasing availability of commercial chemometrics software packages implementing easily used versions of PLS and PCR algorithms. Applications of multivariate calibration have become more commonplace, and relatively few reviews of methods for multivariate calibration appeared during the last two years. Reviews that did appear focused more on the subtle concepts raised in the development of the calibration model than on the mechanics of calibration. Thomas (F1) provided a tutorial review of the logic and statistical underpinning of multivariate calibration. Danzer (F2) offered a perspective on the philosophy of multivariate calibration, qualitative analysis, and their relation in another review. Cerda et al. examined multicomponent analysis of analytes in environmental samples using different analytical techniques (F3). Yu and Xie (F4) reviewed robust methods for multivariate calibration, including robust Kalman filters, robust PCR, and projection/pursuit regression. Improving the methods for multivariate calibration themselves is an active area of research in chemometrics. Several groups offered new PLS algorithms. Cummins and Andrews introduced iteratively reweighted PLS as a robust method for calibration and demonstrated its resistance to the effects of outliers with a Monte Carlo study (F5). Making PLS more efficient when large numbers of variables are present in the x-block has also received attention. An iterative version of the PLS algorithm was reported to be faster and less memory intensive than PLS implemented with the NIPALS algorithm (F6). Kernel-based PLS algorithms also improve PLS performance with large numbers of variables and objects. Ra¨nnar et al. reported new kernel-based PLS algorithms for situations with large number of variables (F7) and large numbers of objects (F8). Cross-validation using a kernel-based PLS algorithm was demonstrated on image data (F9). In addition to speeding PLS, efforts have been made to improve the predictions from PLS by selecting highly informative variables for the calibration model through an interactive procedure (F10, F11).
Improvements in multivariate calibration based on PCR also were reported . Thomas demonstrated the improvement in prediction using PCR by implementing a method for incorporating variation present in instrumental measurements made on the prediction set (F12). Correlation principal component regression was also demonstrated to produce better performing calibration models than conventional PCR (F13). Both PLS and PCR are now commonly used in situations far removed from what their inventors imagined, and the theoretical implications of this tendency are being examined. Research into the structure of PLS continues to refine our understanding of the workings and interrelation of the popular methods for multivariate calibration. Ho¨skuldsson (F14) discussed the relation of PCA and PLS. Langsrud and Næs (F15) demonstrated that, for orthogonal x-block variables, PLS, SIMPLS, and reduced rank regression (RRR) give identical results. de Jong continued his theoretical analysis of PLS in its many forms, offering comments on the new PLS kernel algorithm (F16) and a brief discussion of the shrinkage properties of PLS (F17). Ridge regression, a technique more in favor at present with statisticians than chemists, also attracted attention from those interested in the theory of calibration methods. It was demonstrated that the close relation shown to PLS and OLS regression by ridge regression could be extended to PCR if the ridge parameter is taken as negative (F18). Locally weighted regression (LWR) is another method commonly used in statistics, but it is occasionally used in chemometrics as an alternative to biased regression based on PLS or PCR. A new approach for distance measurement, balancing information in the x- and y-blocks, was demonstrated to improve the robustness of LWR to noise in both the x- and y-blocks (F19). LWR was also shown to produce better predictive models than other methods in near-IR monitoring of fermentation broths (F20). During the last two years, fundamental work on calibration began to be refined in a number of aspects. Special focus was given to outlier detection, interferences, model error, and practical calibration in time-varying systems. Robustness and an ability to deal with imprecisely specified models were issues receiving special scrutiny. In a particularly noteworthy paper, DiFoggio examined several misconceptions in the current practice of chemometric calibration as practiced in near-IR spectroscopy (F21). Using both real and synthetic data, he presented counter examples to demonstrate how the misconceptions concerning the capabilities of chemometric calibration methods arise. Several authors addressed novel improvements to the venerable standard addition method. One paper reported a new algorithm which takes into account multiplicative and additive effects from the sample matrix (F22). The philosophy and mechanics of standard additions, generalized standard addition (GSAM), and dilution standards addition (DSAM) was the subject of a recent review (F23). Multiple standards addition with latent variables was proposed to deal with multiple equilibria in the sample matrix (F24). Three papers appeared on the use of generalized standard addition in specific applications. In one, GSAM was used with PLS for calibration of a flow injection analysis (F25). The two other applications of GSAM dealt with inductively coupled mass plasma mass spectrometry. One reported an extension of GSAM, coupling it to rank annihilation to deal with unexpected interferences (F26) and the other used a simplified GSAM to generate a response surface (F27). Several papers appeared on a method for removal of an unknown matrix absorbance in calibration known
as H-point standard additions (HPSAM). A generalized H-point standard additions was reported (F28). HPSAM showed lower detection limits than derivative spectrometry in determinations made in the presence of unknown backgrounds (F29). HPSAM was applied to synchronous fluorescence (F30) and to the resolution of binary mixtures using multiple standards addition (F31) and HPSAM theory was extended to identify regions of linear response/concentration regions (F32). Analyte/matrix interferences are not the only kind of interference that can arise in calibration, however. Interferences between membership functions in fuzzy regression of olive oils to sensory properties were reduced with a fuzzy filter (F33). Many studies were concerned with the detection of outliers or the reduction of the influence of undetected outliers on the calibration model. Robust regression was used to address partial nonlinearity in a calibration (F34). A comparison of semirobust and robust PLS methods appeared (F35). Brown reported on graphical diagnostics for linearity and interaction between components in a calibration (F36, F37). Statistical methods for determining linearity were also summarized (F38), and a tracking algorithm for nonlinearity in a calibration was reported (F39). Oman and Næs reported a new method for detecting and compensating for nonlinearities in a calibration using principal component regression (F40). Robust regression for detection of outliers was reported by Walczak and Massart (F41). Outliers can result from changes in the analyte matrix between calibration and prediction. A new calibration can be done only if the constituents can be identified, which is the goal of another study on outliers (F42). Walczak reported on detection of outliers in multivariate calibration (F43) and bilinear calibration (F44). Occasionally, errors in both axes must be considered in the calibration. Kalantar et al. discussed bias in the summary statistics of slopes and intercepts in linear regression done on data with errors in both variables (F45). Danzer et al. evaluated several methods for orthogonal least squares, including PCR, and found that when x- and nonnegligible y-block errors are present, orthogonal regression is preferable to ordinary least-squares regression (F46). Model error resulting from incorrectly specified error distributions or model form have been addressed by robust and fuzzy regression (F47). Robust regression and outlier detection for nonlinear models have been implemented using genetic algorithms (F48). A number of papers exploring other calibration methods also appeared. In addition to a very large number of papers making use of simple variants of linear least-squares regression that are not cited here, there were several that will be of interest to the chemometrics community. One especially interesting one concerned Procrustes regression methods. Procrustes regression involves the fitting of one matrix to another matrix observed on the same objects. Unconstrained Procrustes regression was used to reconstruct a mid-IR spectrum from an near-IR spectrum of the same materials. Peaks attributed to the same function groups were identified in both spectra (F49). Inverse least squares (ILS) was used to determine phenol by HPLC (F50) and to speciate arsenic compounds measured by atomic absorption (F51). Dynamic calibration, where reference samples are measured at some interval and weighted regressions are used to update the calibration model, was implemented for an on-line calibration of steel (F52), and the advantages of the extended Kalman filter for compensating for drift in on-line calibrations was revisited in a second report (F53). Reports on calibrations using the Kalman Analytical Chemistry, Vol. 68, No. 12, June 15, 1996
filter for ultraviolet/visible spectrophotometry (F54, F55) and for line atomic spectroscopy (F56) also appeared. Several authors considered the relative merits of several different methods for calibration. In one study, PLS, ILS, CLS, PCR, and the Kalman filter were compared for calibration and wavelength selection in visible spectrophotometric analysis (F57). Another compared OLS with variable selection, PCR, and PLS for process analysis (F58). Several PCR techniques were compared with PLS and OLS for near-IR analysis of pharmaceuticals, where it was demonstrated that variable selection and selection of principal components greatly improved results from PCR calibration (F59). A new method known as restricted PCR, PLS, and two forms of conventional PCR were found to perform similarly on near-IR spectra (F60). These comparisons probed the overall quality of a multivariate calibration by exploring the relatively small differences in predictive power that arise between calibration methods. Keller et al. (F61) also considered the quality of a multivariate calibration, but they used knowledge of noise in the data, the spectra of the analytes, the concentration range spanned, and the number of calibration samples in a Monte Carlo simulation to estimate the predictive quality of the calibration model in advance. In the absence of general theory propagating error through a multivariate calibration, this approach has a good deal to recommend it. Geladi (F62) reported on a model comparison plot to assess candidate models. Forina et al. (F63) evaluated the effectiveness of three validation methods in generating the best predictive model. They found that the commonly used single evaluation set produced poor estimates of the residual standard deviation and the complexity of the PLS calibration model. Crossvalidation gave better estimates when at least 10 cancellation groups were used. Grung and Kvalheim (F64) discussed the effect of sample replication and true instrumental resolution on the number of significant latent variables obtained by crossvalidation. Achievement of a satisfactory multivariate calibration model may not be the final step in many practical applications. Once it is developed, it is often necessary to transfer the calibration model to other instruments, so that the calibration can be used at the point of measurement rather than in the research laboratory. One way to achieve transfer of calibration is to standardize either the instrumentation used or the calibration itself. de Noord (F65) reported on the use of standardization of calibrations in a review, while Workman and Coates discussed a strategy for the standardization of analytical instrumentation (F66). Bouveresse et al. used three standardization sets and Shenk’s algorithm to transfer near-IR calibration models (F67). The PDS algorithm for transfer of calibrations continues to be refined. The latest report concerned the correction of additive backgrounds in transfer of spectra between instruments (F68). Other approaches receiving attention during the last two years included making calibrations from data collected on several instruments known as “co-masters” (F69) and using multivariate control charts to track statistical fluctuations in calibration and prediction (F70). Two PLS regressions, one between spectra on the master and slave and the second on concentrations, have been used to transfer calibrations in near-IR spectrometry (F71). Use of first-derivative spectra to help deal with the small differences in wavelength that arise from instrument to instrument is the subject of another recent report on transfer of calibrations (F72). Papers discussing 38R
Analytical Chemistry, Vol. 68, No. 12, June 15, 1996
issues concerned with transfer of calibrations in polymer production processes (F73) and brewing (F74) also appeared. Higher-order calibrations continue to be explored, and a few analytical applications have begun to appear. Faber et al. reported on bias and variance in the eigenvalues of GRAM (F75). The effects of correlated and uncorrelated error on GRAM was the subject of a paper by Booksh and Kowalski (F76). Data acquisition parameters affecting the accuracy and precision of GRAM and adaptive Kalman filters were evaluated in another report (F77). Iterative rank annihilation, coupled to self-modeling methods, was demonstrated for decomposition of pyrolysis mass spectra (F78). GRAM is just one of the possible second-order calibration methods, however. Smilde et al. discussed the theory of medium- and high-rank second-order calibration (F79) in the context of GRAM and nonlinear residual bilinearization analysis (NRBA). The equivalency of yet another second-order method, residual bilinearization (RBL), and NBRA was demonstrated (F80). Standards additions has also been used for second-order calibration problems. Xie et al. used GSAM with RBL to compensate both matrix background and the influence of unknown interferences in a multivariate calibration (F81). An algorithm for second-order standards addition was reported and applied to the calibration of a kinetic sensor for trichloroethylene (F82). The accuracy of the estimates was verified by Monte Carlo simulations. Trilinear decomposition and three-way methods have begun to appear in practical applications. Second-order tensorial calibration of a heavy metal sensor, using GRAM and trilinear decomposition, was reported (F83). The trilinear decomposition method was applied to a flow sensor (F84). Improvements in the trilinear decomposition algorithm produced more reliable estimates. Three-way PLS was applied to the analysis of spectral data in a kinetic analysis. Results were superior to conventional PLS, especially when small amounts of analyte and small spectral differences in species were present (F85). Applications of multivariate calibration dominated the literature. Use of PLS regression now has become commonplace in analytical chemistry, and applications are appearing in very distant fields, especially for calibration of properties to responses. No doubt, a major part of the increase in the use of PLS can be attributed to improving commercial software for chemometrics, but better education of chemists in the use and application of multivariate calibration also appears to have a role. Vibrational spectroscopy has long been an area where chemometric methods are embraced, and it is no surprise that many of the applications appeared in analyses using near-IR and mid-IR spectroscopy. Real-world effects in multivariate calibrations of near-IR spectra were briefly discussed (F86), and the advantages and disadvantages of the use of higher-order principal components in PCR applied to nearIR data was presented (F87). Analysis of glucose in various aqueous media has been the subject of attention from many groups. Quantitative analysis of glucose in syrups by mid-IR using attenuated total reflection was discussed (F88). The lower levels of glucose found in physiological samples and the high background presents a special challenge to the spectroscopy and the multivariate calibration. Analysis of plasma glucose (F89) and the errors present in predictions from these multivariate calibrations (F90) were discussed by Heise and Bittner. Problems of biocompatibility of components of the plasma sample with the sampling system in mid-IR spectroscopy and their effect on subsequent multicomponent analysis were discussed (F91).
Glucose was determined by mid-IR spectroscopy at physiological levels in simulated serum by multivariate calibration with PLS and ANN (F92). Other common constituents of serum also were targets of multivariate analysis based on mid-IR spectra, including glutamine (F93), urea (F94), cholesterol (F94, F95), uric acid (F95), and albumins, globulins, and proteins (F96). Multivariate calibrations of near-IR spectra for glucose and other analytes of physiological importance have also appeared, including a paper on temperature-insensitive multivariate calibration for glucose (F97), a paper considering the effects of data pretreatment on multivariate calibration of serum glucose measured noninvasively by diffuse reflectance near-IR (F98), multivariate calibration based on an unusual experimental design for in vitro monitoring of glucose and other analytes by near-IR (F99), and calibration of cholesterol, protein, and triglycerides in plasma (F100). Urea was determined in dried serum (F101) and proteins and lipids in serum and whole blood (F102) using near-IR and multivariate calibration methods. As noted above, there were too many applications of multivariate calibration to cite here. Representatives of those publications are summarized in the following paragraph, however. These were selected on the basis of unusual calibration, preprocessing, or unusual measurement systems. Spectroscopic measurements were most commonly used for calibrations, and near-IR spectroscopy was by far the spectroscopic method of choice. PLS was the calibration method of choice in most of these applications, but there were a few interesting exceptions. OLS was used to relate near-IR spectra to nutrient levels in fermentation bioprocesses (F103). PLS was used in several unusual calibrations, including one for polymorphic crystal forms of cimetidine by Raman spectroscopy (F104) and the cis and trans content of fats and oils by mid-IR spectroscopy (F105). Nondestructive analyses by spectroscopic methods and PLS was also popular. Near-IR reflectance and PLS calibration was used to measure moisture content of mushrooms (F106) and the fat, moisture, and protein content of whole salmon fillets (F107), while near-IR transmittance was used to monitor the protein content of single, whole wheat kernels (F108). For many real calibrations, nonlinear effects had to be addressed. A linearized PLS calibration method was used for CO analysis from mid-IR spectra (F109), and linearized data from diffuse reflectance IR were used to calibrate constituent minerals in cements (F110). Locally linear PLS calibration of the diffuse reflectance IR spectra of soils with amounts of metal oxides and other chemical constituents present was found to give acceptable results (F111, F112). IR characterization of thin films with the help of multivariate calibration was reviewed (F113), and it was demonstrated that IR determinations of film properties and composition could be more precise than the reference methods used to establish the PLS calibration (F114). Raman spectra of dissolved gases in aqueous media were used to calibrate the amounts (F115). Ultraviolet and near-IR spectrometry were used to calibrate constituent amounts and particle size properties of a polymer latex (F116). Solid pharmaceuticals measured by diffuse reflectance near-IR were calibrated to reference spectra stored in a spectral library using PLS (F117). Relative precisions obtained from five different methods for multivariate calibration for metal ions measured by diode-array spectrometry in model process streams showed stepwise regression to be superior to other methods (F118).
Multivariate calibration is also beginning to appear in areas long dominated by univariate calibration approaches and involved chemical separations. Yan et al. demonstrated a polymer-based reflectance sensor used to detect binary mixtures of organic pollutants in water. The transient signal profile was successfully calibrated to analyte amount by PLS (F119). Another kinetic method, this one based on stopped-flow spectrophotometry in flow injection, was used with PLS to calibrate two- and three-component systems containing metal ions. This method required no assumptions about the kinetics involved (F120). Laser-induced fluorescence, with a Fourier transform compression and preprocessing step, was used to perform multicomponent analysis of polycyclic aromatic hydrocarbons using PCR calibration (F121). Multivariate calibration in spark spectroscopy and inductively coupled plasma spectroscopy have been discussed (F122). Procrustes rotation and PLS were used to calibrate components measured by Fourier transform pulsed gradient spin echo NMR, and PLS was suggested for use in T1 or T2 measurements (F123). Some interesting applications in chromatography included a patent application in which PLS is coupled with measurement of reflective index at several wavelengths to construct a detector capable of detecting and quantitating several co-eluting compounds (F124). Several papers concerned the use of multivariate calibration with gas chromatography using mass spectrometric (GC/MS) detection. One paper reported a method coupling PLS to GC/MS analysis of steroids and metabolites for rapid discrimination of anabolic steroids in athletes (F125). Another paper discussed the relative merits of conventional GC/MS, using univariate linear regression and a PCR method. In the systems studies, the PCR method performed better (F126). More recently, electroanalysis has been another analytical technique benefiting from application of multivariate analysis. A study by Jagner et al. demonstrates that there are significant advantages to be gained by using multivariate calibration in electroanalysis of systems with several interfering components. They were able to determine As by stripping analysis in the presence of multiple interferent species that, with the conventional univariate calibration methods used by most electrochemists, would have rendered the analysis useless (F127). Contrary to the usual practice with electrochemical methods, the analysis with sensor arrays has long made use of multivariate calibration to enhance analytical selectivity. A recent paper compares PLS with neural networks for calibration of volatile organic compounds by a polymer-coated quartz microbalance sensor array. The neural network performed better than PLS for mixtures of more than two compounds (F128). Calibration to properties continues to be an active area of investigation, too. The breadth and ingenuity of applications appearing in this area during the last two years were impressive. Circular dichroism and IR spectrometry were used to calibrate the secondary structure of a protein from its spectra. Standards were generated from X-ray crystallographic analysis (F129). NearIR spectra were calibrated to mid-IR spectra of polymer blends (F130). The relationship between endothelial cell growth and surface properties of plasma-deposited films as monitored by secondary ion mass spectrometry was investigated using PLS calibration (F131). Niemczk’s group continued exploration of the calibration of properties of thin films from their infrared spectra (F132, F133), and the crystallinity of poly(aryl ether) films were calibrated to their polarized IR spectra (F134). Food chemistry was one of the first areas where calibrations to property were Analytical Chemistry, Vol. 68, No. 12, June 15, 1996
applied. Recent papers in this area included a fiber digestability study, where near-IR spectra were correlated to neutral detergent fiber in the diet (F135), and a paper that used multivariate calibration methods to measure white and red wine blends in rose wine production (F136). Methods for calibrating the sensory properties of bell peppers (F137) and nondestructively measuring the composition of pea seeds (F138) were also reported. Rapid measurement of biomass is another task that benefits from multivariate calibration methods. The biomass content of stored hay (F139) was related to mass spectral measurements. The amount of cellulose I and cellulose II determined in fiber samples was correlated to 13C NMR (F140), and 13C NMR was also used with calibration methods to examine the variations in cellulose structures resulting from differences in Kraft processes (F141). Several unusual applications of calibration illustrate the possible utility of calibration to properties in environmental studies. NearIR spectra of soil and sediment samples containing biosynthetic polymers such as cellulose were correlated to microbial activity (F142). Calibration of near-IR spectra was also used to predict the nutritive value of tree and shrub foliage for small ruminants (F143), the chemical composition and energy content of cattle feed (F144), and the phosphorus content and phytase activity in vegetable feedstuffs (F145). PLS calibration has been applied to the prediction of lignin, cellulose, and nitrogen concentration in forest foliage monitored by remote sensing from satellites (F146). Closer to earth, IR spectra were correlated to several measures of pulpwood quality (F147). A few more calibrations of spectra to petroleum properties also appeared. Raman spectra were correlated to petroleum properties (F148) and to octane numbers and Reid vapor pressures of commercial petroleum fuels (F149). Other reports of calibrations to property included one relating near-IR spectra of cotton fibers to strength (F150) and one relating the odor of linoleum to gas chromatographic signature (F151). Research on methods for nonlinear multivariate calibration also experienced substantial growth in the last two years. Papers reporting applications of ANN were mainly responsible for this growth, but other nonlinear calibration methods also attracted attention. Probably the most unusual application was that of Wang et al., in which the neural networks were developed with an analytic transfer function mathematically describing the nonlinearity specific to the chemical system under study. These “ChemNets” worked well in the examples given, but they are limited to use in places where the mathematical form of the nonlinearity is known in advance, very much like conventional nonlinear regression (F152). Næs et al. gave a short overview of neural network methods in multivariate calibration (F153). A overview of nonlinear multivariate calibrations, with an emphasis on the ruggedness of the calibration, was provided by Gemperline (F154). Probably the biggest challenge to beginners using ANN in knowing how to train networks and when to stop training. The difference between overfitting and overtraining of neural networks was the subject of one study (F155). The rarely used counterpropagation learning method was summarized and applied in constructing inverse models (F156). Walczak and Wegscheider combined neural networks with linear methods (F157) and with fuzzy logic (F158) in a pair of papers. Papers on a few other nonlinear methods also appeared. Nonlinear PLS was reported to be useful in determination of mixed metal ions by compleximetric titrations (F159). A few reports of nonlinear Kalman filters surfaced. Extended Kalman filtering and ANN, both improved 40R
Analytical Chemistry, Vol. 68, No. 12, June 15, 1996
by the addition of fuzzy logic, calibrated time-dependent substrate concentrations in production-scale fermentation (F160). A modified extended Kalman filter was investigated in the determination of urea by a catalytic reaction (F161). A new method using logistic regression for multiple estimation of concentrations in immunoassay was tested on simulated data and compared favorably to existing methods (F162). Many papers compared neural networks and alternative methods for nonlinear calibration of concentrations or other properties. Borgaard gave an overview of neural networks and made comparisons between networks, PCR, and PLS in two papers (F163, F164). Comparisons of PLS, B-splines, and two different forms of neural networks on several real-world data sets were reported (F165). Neural networks were found to slightly outperform linear PLS in calibration of carrageenans measured by IR spectrometry (F166), and networks also outperformed OLS regression in calibration of pollutant gases measured by remote sensing using mid-IR spectrometry (F167). The problem of monitoring physiological levels of glucose in serum has also been tackled by application of nonlinear calibration with neural networks. Neural networks and other methods of multivariate calibration were compared for noninvasive determination of glucose in vivo (F168). Glucose in whole blood was calibrated to mid-IR spectra with PLS and ANN methods (F169). An unusual patent application combined neural network-based multivariate calibration with high-frequency rf spectroscopy for noninvasive monitoring of glucose in vivo (F170). Pharmacodynamics and pharmacokinetics are challenging problems where getting the form of the underlying nonlinear model is often not simple. Neural networks require no such information, and a recent report shows acceptable performance by a neural network in predicting time-dependent concentrations (F171). Several papers appeared where neural networks were applied to the calibration of chemical species probed in complex biological matrices by pyrolysis mass spectrometry. These included measurement of recombinant mammalian cytochrome b5 in Escherichia coli (F172), calibration of mixtures of lysozyme, glycogen, and DNA or RNA and bacterial mixtures (F173), calibration of ampicillin in fermentor broths (F174), calibration of amino acids in glycogen (F175), and calibration of penicillins in strains of Penicillum chrysogenum and ampicillin in E. coli (F176). Neural networks are increasingly finding application in calibration of chemical species measured by sensor arrays. Dynamic responses from polymer-based gas sensor arrays were used in calibration of organic vapors in air (F177). The neural network was trained to provide qualitative and quantitative information from the sensor array responses and was compared to PLS (F178). Networks were used with inputs from arrays of chemically modified, sintered tin oxide sensors to calibrate for hydrogen, methane, CO, and CO2 in simple gas mixtures (F179). Quantitative information from ion mobility spectra of mixtures of bromine, DMF, and HF using an artificial neural network has been reported by Boger and Karpas (F180). Neural network-based multivariate calibration has also been reported for data sets from atomic absorption spectrometry (F181), visible spectrometry of dye mixtures (F182, F183), and chromium tanning solutions (F184). Nonlinear calibration to properties, mainly using neural networks to establish the correlations, has also begun to appear. Neural networks have been employed to correlate octane number and concentrations of various component species to gas chro-
matographic responses (F185). A recent patent application reports use of a neural network for predicting physical properties of hydrocarbons such as gasolines and diesel fuels from near-IR spectra (F186). Biomass estimation using neural networks has also been reported (F187). Total color differences have been correlated to pigment concentration by back-propagation and counterpropagation neural networks (F188). A Kohonen neural network was used to aid in the experimental design used in building the network models for this study. The shelf life of milk was calibrated to measurements from headspace gas chromatography using a neural network (F189), which was found to predict better than a model developed using PCR on the same data. A neural network was found to outperform library searching and visual analysis of band intensities in semiquantitative analysis of urinary calculi (F190). Finally, a patent reported that genetic algorithms were used to select sensors in an array for robust calibration of hydrogen peroxide in solutions of varying pH and temperature (F191). PARAMETER ESTIMATION Most of the papers referenced in this section deal with the mathematical modeling of chemical properties, and some others are dedicated to spectral curve-fitting methods. Recently, a review by Cumpson included a comparison of depth-profile reconstruction methods in XPS and AES (G1). Another paper reviewed different modeling and curve-fitting approaches in kinetics-based singlecomponent determination of noncatalysts (G2). Several methods were compared for the modeling of conductance measurements in nonbrine water samples, in particular MLR, PLS regression, PCR, continuum regression (CR), and ANN (G3). Carlin et al. compared PLS, back-propagation multilayer perception ANN, radial basis function ANN, and an adaptive B-spline modeling algorithm applied to real and simulated nonlinear modeling problems (G4). Roy compared conventional and bootstrap methods for the nonlinear modeling of insecticide toxicity, catalytic chemical reactions, and Michaelis-Menten kinetics (G5). Diverse mathematical models for curve fitting of chromatographic data were evaluated by a Spanish group (G6). The development of EXAFIT, a curve-fitting program for EXAFS was also reported (G7). Very few novel approaches for parameter estimation appeared in the last two years. The bulk of the papers reported applications of LS and modified LS methods. A theoretical article illustrated with examples discussed the issue of weighting experimental data in curve fitting, and the mode of incorporating weights in linear LS was shown (G8). In NMR, several articles mentioned the use of LS methods for modeling or curve fitting (G9-G13). Theory and applications of LS for XPS peak parameter estimation were extensively discussed by Leclerc and Pireaux in three consecutive papers (G14-G16). For the determination of equilibrium constants from titration experiments, it was reported that the LS fit to raw pH data suffered from nonrandomly distributed modeling and measurement errors. Fitting the point-to-point changes in pH proved to be less sensitive to such errors (G17). In another titration study, a robust regression by least median squares (LMS) provided an objective criterion for the determination of the end point (G18). The application of constrained LS minimization in conjunction with the expectation/minimization principle permitted the reconstruction of inverse depth profile from laser photopyroelectric effect impulse response (G19). The parameter estimation
of immunoradiometric assay data was achieved by a generalized LS algorithm (G20). Curve fitting on XPS-like spectra was performed by using an iterative refinement of nonlinear least squares (NLS) with constraints (G21). The determination of spectroscopic parameters was improved by an extension of the NLS curve-fitting technique. This refinement allowed the simultaneous fitting of multiple spectra with fewer fitted parameters than with one spectrum at a time (G22). In kinetics studies, activation enthalpies, entropies, and volumes were fitted by NLS in the frame of a second-order globalization procedure (G23). Koons et al. applied NLS routines based on the simplex and Levenberg-Marquardt algorithms for the extraction of physical parameters from solid-state NMR powder line shapes (G24). In ac-impedance spectroscopy, the hydrogen diffusivity in a TiO2 film and kinetic parameters related to hydrogen adsorption and absorption reactions were obtained using complex NLS fitting method (G25). For modeling, latent variables-based techniques like PCA, PCR, and PLS remain popular among analytical chemists. Wold presented a generalization of PCA and PLS to dynamically updated models, based on exponentially weighted observations (G26). An algorithm was proposed that was able to carry out both PCA and PLS, thus allowing the treatment of problems that involved combination of regression models and variance-based models (G27). In AES, after a line shape analysis performed with the help of FA, the distortions of the peak shape could be eliminated (G28). Gonza´lez and Gonza´lez-Arjona discussed the advantages of target factor analysis (TFA) against MLR methods for checking model equations in linear free energy relationships (G29), and in another study, TFA yielded empirical correlations between capacity factors and partition constants in reversed-phase HPLC (G30). Predictive relationships between GC profiles and variables determined by different methods were found using PLS in an organic air pollution evaluation (G31). PLS also permitted the correlation of surface properties of common polymeric materials to fibrinogen retention on the surfaces (G32). Another approach often mentioned in parameter estimation is the Bayesian statistical method. It was applied to the estimation of NMR spectral parameters (G33, G34) and the extraction of structure/factor amplitudes from powder diffraction data with highly correlated positive and negative intensities (G35). Joergensen and Pedersen presented a method for the numerical assessment of the variability of physical model parameters, using the Bayesian approach (G36). An American group created an algorithm called BAMBAM (BAyesian Model Building Algorithm in Multidimensions) that could extrapolate an NMR multidimensional free induction decay signal, thus improving the resolution of spectral frequencies (G37). The same authors applied the ML principle to the estimation of parameters from multidimensional time domain NMR spectra (G38). In high-resolution γ-ray spectroscopy, the statistical structure of gross and continuum measurements was modeled by application of the ML principle in a nondestructive assay (G39). The problem of model structure determination and parameter estimation in quantitative IR spectroscopy was addressed, and a method was proposed based on ML parameter estimation (G40). A paper described the analysis of the heterogeneous rate of dissociation of Cu(II) from humic substances with regularized LS and expectation/minimization. Both methods yielded different degrees of heterogeneity, and the Analytical Chemistry, Vol. 68, No. 12, June 15, 1996
optimal solution was determined by application of the maximum entropy criterion (G41). Since the last review, a considerable increase has been observed in the application of techniques derived from artificial intelligence (AI) for parameter estimation and parameter optimization. Zupan et al. discussed the theory of counterpropagation neural networks for complex and inverse models and showed some examples of application (G42). The fitting ability of an ANN was investigated for the optimization of HPLC mobile-phase parameters (G43) and for the estimation of kinetic analytical parameters of first-order reactions (G44). In process control, an ANN was used for prediction of feedwater flow rates in a nuclear power plant, using reference instrument readings as inputs (G45). An Austrian group studied the modeling of chemical data by combination of linear and ANN methods. They compared the performances of PCR, PLS, ANN, ANN using PCA scores as inputs, and ANN using PLS residuals as inputs. The data analyzed consisted of simulated, near-IR, and QSAR data (G46). Miyashita et al. compared the performance of ANN, PLS, and quadratic PLS for the nonlinear modeling of 13C NMR shifts (G47). Another AI-based approach consisted of building a predictive model of reaction rate constants using gas-phase mid-IR library spectra and a rule-building expert system based on the mathematical theory of rough sets . The primary objectives of the rule-building process were the discovery and description of structural set relationships. The model was reported to have a better predictive power than a previous model obtained by MLR (G48). Wu and Bellgardt developed a model-based fault detection system for flow injection analysis operations, based on the estimation, filtering, and evaluation of the model parameters. They used the recursive fixed memory method (RFM) for parameter estimation and separated the faults with respect to their dynamics by applying high-pass and low-pass filters to the estimated parameters (G49). Genetic algorithms (GA), another form of AI, were used for the evolutionary determination of physicochemical parameters and concentrations of analytes from photometric titration data. GA proved to be useful for the estimation of starting values of parameters, prior to refinement by another optimization method (G50). Another source of papers is the application of Kalman filtering to modeling and curve fitting. An extended Kalman filter (EKF) was used for estimating the molecular weight distribution in a methyl methacrylate polymerization process, using a detailed polymerization model and various reaction parameter measurements (G51). In studies of consecutive first-order reactions, an EKF was used for the determination of the initial concentration of a reagent by kinetic detection of the intermediate product (G52). The EKF was also applied to the curve fitting of exponentially modified Gaussian chromatographic profiles (G53), and a novel algorithm, called fading Kalman filter, was described for parameter estimation and peak resolution (G54). Some theoretical articles about the mathematics of modeling and curve fitting were published in the last two years. Losev discussed the advantages of describing an asymmetric spectral peak by a simple mathematical function with a minimum number of parameters (G55). The shape of chromatographic peak profiles was analyzed by a numerical integration method to determine their statistical moments (G56). Parczewski and Kateman suggested an approach for prediction and visualization of correlations between parameters of the mathematical models of processes, by fitting the models to experimental data or finding the extremes 42R
Analytical Chemistry, Vol. 68, No. 12, June 15, 1996
of a fitting criterion. This approach was said to be particularly useful for the determination of parameters in nonlinear models (G57). A procedure was presented for fitting nonlinear working curves in cases where the rate of introduction into an instrument differed in an unknown manner from one standard to the next. The method implied a simplex optimization of the values of the rates, after generating a set of nonlinear simultaneous equations (G58). A new method was investigated for analyzing multicomponent NMR relaxation signals. The idea was fitting a noisy spectrum to a sum of weighted exponentials which number was determined by examination of the SVD of the signal (G59). The SVD was also performed for parameter estimation of NMR signals previously enhanced by a new algorithm based on the minimum variance estimation method. The algorithm was reported to offer better resolution and convergence than nonenhanced methods (G60). Phillips et al. described a new algorithm of nonlinear curve fitting for reduction of optical remote sensor FT data. This algorithm allowed the detection of parameters not readily determined by linear methods, like classical LS (G61). Majer tested newly developed mathematical functions for the modeling of sigmoidal sensitometric curves representing quantitative characteristics of photographic materials (G62). A French group successfully applied Procrustean analysis, a matrix-fitting technique, for the reconstruction of mid-IR spectra of commercial oils from their corresponding near-IR spectra (G63). STRUCTURE/ACTIVITY RELATIONSHIPS The use of multivariate methods in building models that relate molecular structure to a physical, chemical, or biological property is reviewed in this section. Also included in this section are studies that describe the relationship between material composition and property. The relationship between a property and an experimentally measurable quantity such as a spectrum is covered in the section on calibration or pattern recognition. Even with this arbitrary distinction, there were a large number of citations found in the Chemical Abstracts database during the review period. Most research in structure/activity focused on the development of a linear or nonlinear regression model that related multiple electronic or topological molecular descriptors as independent variables to a dependent property variable. Hence, it should come as no surprise that research in the area of descriptor development was an active area during the most recent reporting period. Randic et al. (H1) developed an algorithm for construction of new topological descriptors. The algorithm allows the user to include molecular properties as source data for the construction of novel descriptors and should double the number of molecular descriptors available to chemometricians in their quest for novel property/activity relationships. Log P or other measures of hydrophobicity that often play an important role in structure/ activity relationships (SAR) could be estimated by semiempirical methods (H2), micellar liquid chromatography (H3, H4), reversedphase liquid chromatography (H5-H7), or paper chromatography (H8). Comparative molecular field analysis (H9-H12), which is based on the analysis of steric and electrostatic fields of molecules mapped by a probe atom in a molecular mechanics force field, was utilized in several studies to develop descriptors for threedimensional SAR. Principal properties, descriptors derived from a large set of physical and chemical parameters by PCA, have been utilized in SAR studies involving biologically active molecules
(H13, H14). NMR, a useful source of electronic and structural descriptors, is, at last, being exploited for this purpose in structure/activity studies (H150, H16). During this reporting period, PLS has emerged as the method of choice for data analysis in three-dimensional QSAR, in large measure, because of an advanced variable selection procedure called GOLPE (Generating Optimal Linear PLS Estimations), which is aimed at obtaining optimum PLS regression models (H17-H20). GOLPE employs a D-optimal design in the loading space to implement variable preselection followed by an iterative evaluation of the importance of the individual variables on model performance. The combination of experimental statistical design and PLS has also proved beneficial in many other SAR applications involving PLS (H21-H23). One of the most active areas in structure/activity modeling over the past two years concerned the relation of chemical structure to biological activity. Several reviews on the role of chemometric methods in molecular design (H24-H32) have been published including a multiauthored book edited by van de Waterbeemd (H24). Artificial neural nets, a class of highly flexible nonlinear regression methods, have been examined for modeling effectiveness by Geladi (H33), who used neural nets to predict the acute toxicity of 38 organic compounds with diverse chemical structures. Geladi observed in his study that nonlinear methods, whether derived from PLS regression or back-propagation neural networks, did better than linear methods for describing the relation between acute toxicity and molecular structure. Backpropagation, in turn, did better than nonlinear PLS. The most active area in structure/activity modeling over the past two years concerned the correlation and quantitative prediction of retention from chemical structure. Several review articles have appeared in the chemical literature on structure/retention relationships during the reporting period (H34-H36). Analysis of biochromatographic data obtained from systems incorporating biomacromolecules, or bioactive chemical entities, e.g., affinity chromatography, RPLC, or charge-transfer TLC, may provide information that can be utilized in the design of combinatorial libraries (H37-H52). The use of theoretically calculated indexes or properties as predictors for chromatographic retention continues to be a very active research area (H53-H64). However, questions remain about the efficacy of this approach for prediction of retention. The use of structure/activity relationships to confirm retention mechanism theories is another research area of active interest within structure/retention modeling (H65-H73). Although the correlations developed in these studies are often true, they usually are of a general nature and cannot help in arriving at final conclusions that are chromatographically relevant. On the other hand, multivariate characterization of HPLC mobile or stationary phases has proven to be useful in identifying mobile or stationary phases with similar or dissimilar properties. The selection of retention probes is important in these studies since the types of retention probes used define the specific physical or chemical interactions of the solute with the mobile or stationary phases. Using this approach, selectivity in thin-layer and reversedphase liquid chromatographic systems has been studied by PCA, hierarchical clustering, and canonical variates (H74-H78). Olsen and Sullivan (H79) demonstrated that C-18 bonded-phase silica columns could be categorized in a meaningful way using chromatographic test mixtures that were reported in the literature to probe hydrophobicity, free silanol interactions, metal activity, and
shape selectivity of reversed-phase columns. They used PCA and cluster analysis to analyze the data. Massart et al. (H80) used spectral map analysis to analyze retention data obtained under specific chromatographic conditions for different batches, brands, and types of RPLC stationary phases. He was able to identify stationary phases with similar and dissimilar properties using nine benzene derivatives as retention probes and demonstrated that only three of the compounds were necessary to fully characterize the set of commercial stationary phases studied. Altomare (H81) used comparative molecular field analysis and PLS to develop a quantitative structure/enantioselective retention relationship which yielded some insight into the physicochemical properties primarily responsible for chiral recognition on a DACH-DNB chiral stationary phase. Nurok et al. (H82) constructed regression models that described the dependence of retention and separation quality in planar chromatography to mobile-phase properties. Linear solvation free energy relationships have also been utilized to characterize stationary-phase interactions in RPLC, as well as GC. Model equations developed using either multiple linear regression or factor analysis (H83) have been employed to investigate the structural properties governing retention mechanisms in GC and HPLC stationary phases. Altomare et al. (H84) investigated the potential of C-8 bonded phases to model lipophilicity using linear solvation energy relationships. Park et al. (H85) used linear solvation energy relationships to choose the best column for a given separation among a number of nominally equivalent columns. Poole et al. (H86-H88) used Abraham solvation parameters and PCA to characterize GC stationary phases, including 38 liquid organic salts. These same workers (H89) also used linear solvation energy relationships to investigate the influence of temperature on the mechanism by which compounds are retained in gas/liquid chromatography. A large number of researchers have attempted to develop models that can predict some type of spectroscopic response from chemical structure. The focus of much of this work has been 13C NMR spectral simulation. Clouser and Jurs (H90) used regression analysis and neural nets to simulate directly the 13C NMR spectra of tetrahydropyrans. However, the results of the simulations were extremely sensitive to molecular geometry requiring the use of a Boltzmann-weighted averaging method for manipulation of geometric descriptors. Svozil (H91) employed a Kohonen network to predict 13C NMR chemical shifts of alkanes using substructure-driven descriptors. Evidently, only topologically based descriptors are necessary for predicting the 13C NMR spectra of simple molecules like alkanes. For more complicated molecules, however, topological descriptors such as graph theoretical indexes will only be of value if they incorporate information about bond angles, dihedral angles, or other three-dimensional molecular features which evidently was the driving force for the development of an algorithm by Balasubramanian (H92) for partitioning nuclei into meaningful classes prior to descriptor development. Graph theoretical constructs have also been used to develop a rigorous deduction for general purpose structural elucidation from multiple-dimensional NMR data (H93), and these constructs also form the basis of a software system called SERENDIPITY (H94) for assignment of secondary protein structure from multidimensional NMR data. Spectra/structure correlations have also been investigated using both Kohonen and counterpropagation neural nets (H95). Counterpropagation neural nets is probably better suited for this type of application since Analytical Chemistry, Vol. 68, No. 12, June 15, 1996
these nets offer the possibility of simulating spectra directly from their structural representations, which is a direct result of how this network is trained. There were also some unusual applications of structure/ property modeling reported in the recent literature. The odor of sausages could be correlated to the composition of their volatiles by PLS (H96). Volatile compounds from corn-based snacks (H97) could be correlated to their odor. Principal component analysis of the volatile data indicated that storage time affected significantly the aroma, flavor, and volatile composition of the snack samples. The relationship between wheat flour protein fraction and the bread-making quality of winter wheat cultivars was investigated using PCA (H98). Finally, PCR was used to the predict shelf life of pasteurized milk from the composition of its volatiles (H99). PATTERN RECOGNITION The overall goal of pattern recognition is classification. Developing a classifier from spectral, chromatographic, or compositional data may be desirable for any number of purposes, including source identification, detection of odorants, presence or absence of a disease in a patient or animal from which a sample has been taken, and food quality testing, to name a few. The classification step is often accomplished using one or several techniques that are now fairly well established, including PCA, hierarchical clustering, k-nearest neighbor (KNN), soft independent modeling by class analogy (SIMCA), and statistical discriminant analysis (SDA). Few novel pattern recognition methods were published during the past two years. Instead, the chemical literature on pattern recognition focused on novel and not-so-novel applications. Nevertheless, classification of data remains an important subject in chemometrics, as evidenced by the large number of citations appearing in the Chemical Abstract database on pattern recognition applications during this recent review period which were rivaled only by calibration. Hence, references in this section are organized according to the type of application, namely, applications to spectroscopy and chromatography, sensors, chemotaxonomy, archeological and forensic chemistry, environmental chemistry, food, polymers and plastics, textiles, and diverse industrial applications. The most novel research in pattern recognition involved work with ANN. Interestingly enough, neural networks were also the most frequently applied pattern recognition techniques for spectral analysis over the past two years. In most cases, a layered, feedforward neural network trained by back-propagation of error was used. Otto (I1) reviewed applications of neural nets in analytical spectroscopy. Smit et al. (I2) noted that drift, which may cause the neural network to misclassify objects when the class clusters lie relatively close to each other, can be corrected using the amount of drift as an extra input variable in the neural network. Radomski et al. (I3) showed that feed-forward neural networks can unambiguously recognize spectra at a signal-to-noise ratio significantly below that needed for by-eye interpretation. Meyer et al. (I4) showed that network architecture can be minimized without a concomitant reduction in prediction performance when the principal component scores of the training and prediction set spectra are input elements for the network. Li and van Espen (I5) observed that neural nets performed better than conventional pattern recognition methods for classification of spectra when network parameters are optimized, e.g., scaling and learning mode, range of the initial weights, and transfer function. Varmuza 44R
Analytical Chemistry, Vol. 68, No. 12, June 15, 1996
et al. (I6) also observed that neural nets perform better than SIMCA, SDA, KNN, and other conventional pattern recognition methods. However, the conclusions drawn by both Varmuza and Li may be considered somewhat speculative since a major difficulty currently confronting researchers in this field is the lack of databases containing high-quality spectra. Tanabe and Uesaka (I7) reported that neural nets can identify IR spectra in a shorter time as compared to ordinary database search methods. Luinge et al. (I8-I10) compared neural nets to PLS for interpretation of IR spectra. In this application, PLS has many of the same attributes as canonical variates. The predictions obtained from both approaches were comparable and are as good as the results obtained by human experts, but the time required for training is considerably shorter for the PLS approach. Levy et al. (I11) reported that neural nets based upon both a supervised and unsupervised paradigm together with feature selection exhibit better capabilities than neural networks based on a single paradigm. Wenke and Kateman (I12) investigated adaptive resonance theory-based ANNs for treatment of spectral pattern recognition problems. Some other interesting applications of feedforward neural networks in spectrochemical analysis include the identification of amino acids from second-derivative near-IR spectral data (I13), identification of neural transmitters from Raman spectra (I14), and automated assignment of NMR spectra of proteins (I15). Pattern recognition methods can also be employed to aid in the characterization and design of chromatographic systems. For example, a fuzzy divisive hierarchical clustering algorithm for the optimal choice of solvent systems in thin-layer chromatography was proposed (I16). The method described is capable of finding the optimal combination of solvent mixtures using the quality partition coefficient as the object function. Betti et al. (I17) analyzed 47 flavonoid compounds by TLC using seven different combinations of mobile and stationary phase and used PCA to identify the minimum number of chromatographic systems (a RPC18 and two normal-phase cyano and silica systems) needed for compound identification. PCA was also used to select and reduce the number of test compounds necessary to detect differences in stationary phases designed for the analysis of basic compounds (I18). Juvancz et al. (I19) also used PCA to characterize several new GC stationary phases. He observed that the first principal component represented selectivity based on polarity, whereas the second component showed Lewis acid/base characteristics of the phases. PCA and correspondence factor analysis were used to investigate the relationship between carboxylate selectivity and the concentration ratio of sodium bicarbonate to carbonate in ion chromatography (I20). Peak purity, which is an important problem in applications of HPLC to pharmaceutics, was also investigated during this reporting period using PCA. Massart et al. (I21) examined the effects of different data-preprocessing methods as it applied to the detection of impure and pure zones in the chromatograms using spectra obtained from the diode array, whereas Kvalheim (I22) utilized experimental design techniques to enhance the performance of his HELP technique. The combination of sensors and pattern recognition continues to be an active area of research. Many research groups have directed their attention toward the development of sensors in which neural networks play a vital role. Dieter et al. (I23) reviewed the current status of SnO2 sensors and noted that significant improvement in the performance of these sensors can
be expected in the future as a result of improved selectivity and drift compensation through application of neural nets and other pattern recognition techniques. Bull et al. (I24) showed that improved transducer specificity can be achieved through incorporation of discrimination algorithms based on ANNs. He reported that better results were obtained when neural networks were used with an array of sensors where each cell in the array operated under different conditions. Hashem et al. (I25) reported that using neural networks to analyze sensor data improved significantly the selectivity of the sensor array, especially when some or all of the sensors are not selective. Di Natale (I26-I28) has reported that hybrid neural networks which exploit the benefits of both feed-forward systems and self-organizing maps are promising new methods for increasing the overall selectivity of sensor arrays. Stetter et al. (I29) has reported that neural nets can outperform KNN when instrumental parameters are taken into account by the net while it is training. Neural networks will not only enhance selectivity but can also improve the sensor response time as well (I30). The use of neural networks with a variety of sensor configurations, including SAW devices (I31, I32), quartz resonators (I33, I34), and electrical conductance (I35), has produced truly impressive results. Other interesting applications of sensors and neural networks include the development of multivariate sensor systems for automatic control of wastewater streams (I36), the use of multiarray polymer sensors for determining odor quality of raw materials used in foods (I37), sensors for beer flavor monitoring (I38), and sensor arrays for identification of the different vintages and aromas of wines (I39, I40). Chemical taxonomy, which is concerned with analyzing the chemical expression of type and condition using family- or organism-specific compounds, has long been an active area of research, and the past two years proved no exception. Many chemical taxonomic studies require the use of high-performance chromatographic methods. Objective analysis of the chromatographic data generated often requires the use of pattern recognition methods. Using KNN, Ramos et al. (I41, I42) was able to differentiate various species of Mycobacterium on the basis of HPLC profiles of their mycolic acid patterns. Using gas chromatography and pattern recognition methods to characterize fatty acid profiles of bacteria, a number of investigators have shown that it is possible to differentiate among closely related species or strains (I43-I49). Pyrolysis mass spectrometry and pattern recognition methods have also proven useful in microbiology. Methods for the rapid identification of streptomycetes (I50) and verocytotoxin production in E. coli (I51), as well as differentiation based on Gram type using fatty acids (I52) could be achieved by pyrolysis mass spectrometry/pattern recognition. Harper et. al. (I53) reported that time-varying pyrolysis mass spectrometry/ pattern recognition could be used to detect and classify air-borne biological agents. The pyrolysis data were projected onto a subspace using singular value decomposition, and a convex cone was then constructed in the subspace, showing as its corners physically meaningful components of the sample. This technique enabled separation of the biological material signal largely independent of the absolute amount of sample, and detection of the presence of any biological material could be accomplished based on the convex cone alone. A few papers focused on the use of pattern recognition in the medical sciences. Brown et al. (I54) used near-IR spectroscopy and PCA to differentiate between normal and malignant pap smear
samples. ICAP emission spectroscopy and pattern recognition methods (I55) were used to differentiate between brain parts obtained from so-called normals and from individuals with Alzheimer’s disease (AD). Although the concentration of 18 elements was determined by ICAP, only two elements, Al and Zn, possessed high discriminatory power, which suggests a possible link between these two elements and AD. Correspondence factor analysis was used to investigate possible links between the geographical distribution of AD cases and the geochemical profile of the SLSJ territory in the province of Quebec (I56). Choo et al. (I57) reported that IR and PCA can be used in the diagnosis of AD from autopsy tissue. There were only a few papers that focused on the application of pattern recognition techniques to forensics, which is surprising in view of the potential impact that multivariate methods can have on this field. Using pyrolysis gas liquid chromatography and PCA, Rietjens et al. (I58) reported that it may be possible to differentiate plastic bond explosives by lot. A portable gas chromatograph that took advantage of neural networks for data reduction was developed by Fox and Hooley (I59) for identification of explosives. Freeman et al. (I60) developed a potential method for DNA fingerprinting which involved pyrolysis mass spectrometry and PCA to distinguish between oligonucleotides differing only minimally in either base content or sequence. A method to classify amphetamine samples according to their impurity content to ensure reliable sourcing was developed by Jonson (I61). Jonson’s method uses gas chromatography and PCA to analyze octane extracts of the tablets. Pyrolysis gas chromatography and cluster analysis were used by Hida et al. (I62) to classify hashish in order to estimate the route through which it had been obtained by the suspects. Hiraoka (I63) showed that trace elements such as Sr and Rb can be used to classify soils, and the classification showed strong agreement with geological features. Pattern recognition methods have become an integral part of environmental studies over the past few years. During the course of this review period, PCA was often used for classification of environmental data, frequently to apportion the contributions of various sources of pollution. One study with that objective was carried out by Hopke et al. (I64), who developed source/receptor relationships for inorganic ionic species which were measured during the southern California Air Quality study of 1987. Mizohata et al. (I65) used PCA and target transformation factor analysis to identify the emission sources of particulate matter together with their chemical component profiles. The results of factor analysis indicated the presence of three particle sources: automotive exhaust, road dirt, and tire and brake dust. Michaued, et al. (I66) applied PCA and hierarchical clustering to compositional data and verified that a relationship exists between the particle’s size and its chemical composition. Mendez et al. (I67) used PCA to analyze PAH data obtained from GC/MS analyses of airborne particulate matter. They were able to correlate the compositional data to the anthropogenic activity and meteorological conditions. Using a modified Kohonen neural network (I68-I70), Wienke et al. were able to classify airborne particles by size and shape from their SEM images. He was also able to identify airborne particle sources by their multielement trace patterns using a Kohonen neural network combined with a Prim’s minimal spanning tree. In water research, PCA and a Procrustean rotation were used to select a subset of analytical variables that were measured in an initial broad study to monitor groundwater quality, in order to Analytical Chemistry, Vol. 68, No. 12, June 15, 1996
reduce costs, effort, and staff workloads (I71). N-Way PCA was used to study time-dependent patterns of amino acid concentrations as determined by HPLC for different lakes of the Berlin area in Germany (I72). PCA was used to study eutrophication of a shallow temperate lake in Italy (I73, I74), in order to verify the associations that exist among variables and to separate factors responsible for the observed increase in eutrophication. Contamination of fish by organochlorine compounds (I75) or PCBs (I76) has been studied by GC/MS/PCA, and unique contamination composition profiles characteristic of possible sources were identified. Lavine et al. (I77) showed that underground fuel spills could be traced to their source using SIMCA and other biased pattern recognition methods. Laser-induced fluorescence emission spectra analyzed by backpropagation neural nets, also yielded profiles characteristic of fuel type (I78). A laser fluorescence system for remote characterization of fuel spills was developed (I79), and PCA of the 31 spectral parameters (eight spectral ratios, and fluorescence lifetime parameters) also indicated patterns in the spectral data characteristic of fuel type. A significant number of references involving pattern recognition focused on applications in food chemistry. These applications were targeted in six distinct areas: (1) wines, (2) olive oil, (3) essential food oils, (4) food gums, (5) coffee, tea, or milk, and (6) fruits. Generally, routine analytical measurements, including GC, LC, and trace metal analysis, were used to obtain the data, with PCA and statistical discriminant analysis constituting the bulk of the multivariate data analysis techniques employed. Applications of pattern recognition methods in studies of food protein functions have been reviewed by Nakai et al. (I80), and Martinez-Anya et al. (I81). Wine applications focused on aging and type. Linear discriminant analysis of HPLC profile data of polyphenolic compounds found in wine (I82, I83) revealed information about aging. PCA of the volatile components of white wines revealed concentration patterns in the data indicative of type (I84-187). Year-to-year variability in white wine composition was investigated using SIMCA, PCA, and canonical variates (I88). Elemental analysis and PCA were used to differentiate white wines by region (I89). Olive oil applications focused on sensory attributes and geographic origin. Odor was correlated with headspace composition using PCA and PLS (I90, I91). Near-IR and PCA were used to differentiate authentic olive oils from adulterated ones (I92). Olive oils have also been classified by region (I93, I94) via PCA of fatty acid compositional data. Essential oil applications focused on differences in fatty acid composition. Mandarin essential oils extracted by different technologies could be distinguished by PCA of their fatty acid profiles (I95). Nine varieties of vegetable oils could be successfully classified from PCA scores of their near-IR spectra (I96). Essential oils could be classified by region using GC/MS/PCA (I97, I98). FAB/PCA (I99) could differentiate edible oils on the basis of differences in their triacylglycerols. FTIR/PCA was used to authenticate vegetable oils and differentiate among various essential oils in lipid-rich foods (I100, I101). Food gum applications focused on authentication. PCA and LDA were used to cluster gums by geographic region or climate (I102I106). This type of analysis can provide a practical method for identifying adulteration or commercial samples that fall outside of agreed limits. Coffee, tea, or milk applications focused on adulteration and geographic origin. PCA of IR spectra have been used as a classification method for differentiating among various species of coffee beans (I107, I108) which is important to ensure 46R
Analytical Chemistry, Vol. 68, No. 12, June 15, 1996
high quality, particularly for ground and roasted coffee products. Classification of coffee cultivars was also implemented, using PCA of either the headspace GC profiles of the volatiles (I109) or HPLC profiles of the chlorogenic acids (I110). Neural networks were used to classify black teas by origin on the basis of differences in their phenolic compositions (I111). Adulteration and flavor quality in milk have been correlated to milk’s fatty acid and volatile composition via PCA (I112, I113). Fruit applications also focused on type and adulteration. Infrared spectroscopy and PCA were used to determine the type of fruit in jam (I114) and to authenticate fruit purees (I115). Orange juice authentication, which continues to be an area of active research, was pursued in a number of different ways including PCA of HPLC flavonoid profiles (I116), SDA of principal components derived from nearIR spectra of fruit juices (I117), and PCA and canonical variates of volatile constituents of orange juice (I118, I119). Pattern recognition techniques continue to be exploited in a wide variety of industrial settings. Using a near-IR camera to sort plastics from nonplastics, Wienke et al. (I120) showed that singular value decomposition can improve the measured multivariate images of a recorded objects, which in turn can be submitted for analysis using a new classification algorithm called multivariate image range analysis. Scott and Waterland (I121) proposed a compact rugged instrument for sorting plastics based on fixed-filter near-IR and neural networks. Alam et al. (I122) has developed an infrared spectrophotometer-based system for sorting consumer plastics which also uses back-propagation neural nets to process near-IR data. An FT-IR-based technique for identification of plastic parts which utilizes reflectance at the surface of the product, with pattern recognition and library searching techniques for data postprocessing, has been developed by Jansen and Hastenberg (I123) and is capable of discriminating between a large number of plastics including black ones. In a textile application, DRIFTS and PCA have been used to differentiate between various types of fabrics, including cellulose (I124) and cotton fabrics (I125), as well as dye mixtures extracted from fibers in worn clothing (I126). In pharmaceutical applications, near-IR and pattern recognition methods have been employed to ensure quality control in the manufacturing process. A review on pattern recognition approaches available to users of near-IR has been published (I127). Cuesta and Sanchez (I128) reported on a method for monitoring powder blends by near-IR spectroscopy which involves DRIFTS in the near-IR range and PCA to compare spectra. Dempster et al. (I129) described an near-IR method for noninvasive identification of film-coated and nonfilm-coated blisterpacked tablets for clinical trial supplies which utilizes SIMCA and the Mahalonobis distance for data postprocessing. Near-IR and SIMCA have also been proposed as an alternative to biological testing for quality control of hyaluronan (I130), a high molecular weight carbohydrate used as a medical device for eye surgery. Gemperline and Boyer (I131) have developed a method for classification of near-IR spectra of tablets based on a sample’s normalized distance from a library of mean spectra. Several researchers reported studies on the development of new pattern recognition techniques or the refinement of established ones. Gieser et al. (I132) claims that a display technique that maps highly correlated data by the use of nonorthogonal vectors is superior to eigen-based methods and is easier to use. As many as 35 chromatographic variables can be simultaneously visualized in a single three-dimensional map using cosines of the
correlation coefficients as vector angles. Currently, there are no studies to support Gieser’s claim. Hartmann and Massart (I133) also developed a display method for visual comparison of the results of two measurement methods based on PCA. The method has been compared to Bland and Altman plots using simulated data. Cho and Gemperline (I134) developed a classification method in near-IR based on a robust distance estimate determined by minimum-volume ellipsoid estimators of multivariate location and scatter. Nierop et al. (I135) developed a variation of discriminant analysis called reflected discriminant analysis which is based on a model that contains many of the attributes of both discriminant analysis and PCA. Wienke and Buydens (I136) investigated the utility of adaptive resonance theory-based neural nets as a means to accomplish real-time pattern recognition in chemical process monitoring. Forina et al. (I137) investigated the viability of double cross-validation as a method to determine the number of significant principal components. LIBRARY SEARCHING Library searching has become an important tool in the identification of unknowns and the qualitative analysis of mixtures. During this review period, only a few citations on library searching systems were found in the Chemical Abstracts database. As in previous years, the literature in this area of chemometrics has been primarily focused on applications. For the purpose of this review, however, the ensuing discussion will emphasize novel procedures or the enhancement of existing methods as well as unusual applications. A number of previously developed library searching methods have been reexamined for the purpose of refinement or adaptation to new applications. Dathe and Otto (J1) utilized a library search method that takes advantage of the dependencies of absorbance on the modifier content of binary solvent mixtures for identifying spectra generated by an HPLC diode-array detector. This new procedure, which was developed to interpolate spectra at every composition of eluent in the model range, utilizes both the Euclidean distance and the correlation coefficient to compare spectra. Brown and co-workers (J2) successfully applied a library mixture search method originally developed for IR spectra to UV/ visible absorbance spectra. The method is based on PCR and adaptive filtering. Wilkins (J3) used a back-propagation neural network as a prefilter to classify library IR spectra on the basis of functional group. The network generated sortable bit string keys for the spectral library which allowed the construction of a binary search tree ensuring successful and fast spectral retrieval. Otto (J4) also utilized a back-propagation neural network to implement a spectral library search system. He showed that a well-trained net can produce results superior to classical methods of library search for noisy spectra or spectra containing impurities. The maximum common substructural algorithm developed by Chen and Robien for searching a 13C NMR database was used to deduce automatically the structural fragments comprising an unknown compound by identifying a set of similar structures in the database via spectral search techniques (J5). Some new techniques were developed for library searching. Gross and Adams (J6) developed a peak-matching algorithm that incorporates fuzzy logic to identify IR spectra by matching them to a standard IR database. A method for estimating the probability of correct identification in a library search using spectral match factors reported by the library search system was developed by
Stein (J7). A chemical substructure identification algorithm (J8) for electron ionization mass spectrometry was also developed by Stein. The algorithm is similar to Isenhour’s K-nearest neighbor method with improvements arising from spectrum screening, peak scaling, and utilization of a better distance weighting scheme and a larger reference library. Testing and optimization of existing spectral library search algorithms was also a topic of interest during this review period. Stein and Scott (J9) compared five library search algorithms for identification of unknown compounds from low-resolution mass spectral data. The best performing algorithm was the dot product function. Other methods in order of performance were the Euclidean distance, absolute value distance, probability-based matching, and Hertz. Boruta and co-workers (J10) compared Euclidean distance to least squares for mid-IR data and observed that limiting the spectral range and/or prefiltering the search algorithm by peak, substructure, or functional groups can markedly improve the performance of a library searching algorithm. The composition of the database can also influence the outcome of a library search. Durindova (J11), Smith (J12), and Stan (J13) have shown that smaller specialized libraries outperform larger commercial libraries and may be necessary to ensure reliable computer-assisted interpretation of spectral data. Finally, some novel or interesting ways to use databases have also appeared. A database of analytical methods for AAS designed by Kateman and co-workers (J14) uses fuzzy logic to select the method that best fits the requirements of an analysis. Employing in situ UV spectroscopy with library searching, Ojanperae (J15) has shown that TLC can be used successfully on an interlaboratory basis to analyze drugs in autopsy specimens. Thomas (J16) has identified a remnant dye on an excavated textile using FT-IR microscopy equipped with a spectral data library of fibers, dyes, and mordants. Building a library of tablet types consisting of second-derivative spectra, Jones (J17) has shown that fiber-optic near-IR can be used to differentiate among different types of blister-packed tablets used in clinical trials. Yates (J18) has shown that fragmentation patterns observed in the tandem mass spectra of peptides containing covalent modifications can be used to directly search and fit linear amino acid sequences present in a database. Laber (J19), Rozenblum (J20), Schuberth (J21), and Fuller (J22) have shown that two-dimensional searching using combined spectral and/or chromatographic libraries is both practical and often necessary in the identification of unknowns in complex mixtures. ARTIFICIAL INTELLIGENCE AI is probably one of the areas of which development has most deeply influenced the approach of chemists to data analysis in the last twenty years. They learned to use computers, not only to produce fast and accurate calculations but also to generate rules, models, and decisions, and even refinements such as the notion of human-like fuzzy reasoning. AI first appeared in chemistry through expert systems (ES) used by engineers for on-line process control. Expert systems are now currently applied to the choice of analytical techniques or conditions classification of spectra, structure elucidation of compounds, and many other tasks where they bring a consistent help to the chemist. More recently, ANNs have emerged as highly flexible tools capable of handling nonlinearities in problems of modeling or pattern recognition, without requiring knowledge of underlying functions or distribuAnalytical Chemistry, Vol. 68, No. 12, June 15, 1996
tions, a tremendous advantage over statistical techniques. This section mentions mainly papers on the theory or improvements of neural computation in chemistry, but other applications can be found in other sections such as Calibration, Structure/Activity Relationships, Resolution, and Parameter Estimation. Fuzzy logic is becoming more and more popular to solve problems where crisp decisions are inappropriate. Fuzzy logic is often coupled to an expert system, thus generating “smart” decision-making systems. GAs, a computation based on the theory of natural evolution. are another form of AI recently developed. GAs are particularly valuable to find optimal solutions when the search space offered to the chemist is too large to be exhaustively investigated. A few papers only in this section mention GAs; most of them appear in the section on Optimization. In 1994, the Third Symposium on Automation, Robotics and Artificial Intelligence Applied to Analytical Chemistry and Laboratory Medicine was held in San Diego. The proceedings of this symposium appeared in Chemometrics and Intelligent Laboratory Systems. Several reviews related to AI were published recently. One of them discussed the use of ES for chromatography (K1). Bell and Mead reviewed the AI and ANN techniques applied to ion mobility mass spectrometry (K2), and some new techniques of machine and discriminant analysis are invoked as alternatives to ANN for protein secondary structure prediction (K3). A concept of inductive reasoning for QSAR intelligent systems, based on empirical risk minimization and the model of structure description, was discussed by Sapegin (K4). An overview of ES in trace analysis was published (K5), and a paper invoked ES for the selection of optimal protein purification processes (K6). Another article was dedicated to the role of AI in analytical systems for the clinical laboratory (K7), and in the second part of their review concerning GA, Lucasius and Kateman compiled concepts and techniques related to GA (K8). Davis published a review on the generation and use of hierarchical trees for spectral analysis (K9). A chemical recognition software was created, consisting of a multivariate analysis package for calibration of complex mixtures, a genetic optimizer for customization of the multivariate algorithm, and an ANN-based filter for the choice of chemicals in a database (K10). In spite of the ever-growing concurrence of ANN in some of their areas of application, ES remain a very attractive form of AI to many chemists, especially because their output is based on historical databases and human knowledge, whereas ANN are still considered as “black boxes”. A first task at which ES can be invaluable is helping the chemist choose instruments and procedures for a given analysis and detect possible faults. In chromatography, several authors reported the use of ES for the determination of the optimum eluent composition (K11-K13), the appropriate detector for ion chromatography, based on inductive reasoning (K14), and other chromatographic conditions (K15). ES combined with hypermedia tools were developed for guiding experimenters using a spectrofluorometer (K16) and for the determination of dissolution methods in AAS (K17). In electrochemistry, Esteban et al. enhanced their ES designed to guide the analyst for the voltammetric detection of trace metals by extending the number of metals that could be detected (K18K20). An ES was developed for the diagnosis of faults in GC data analysis (K21). The problem of knowledge acquisition for a GC diagnosis ES was addressed in another paper, and the knowledge necessary was generated by tracing departures of a specific 48R
Analytical Chemistry, Vol. 68, No. 12, June 15, 1996
component from a theoretical model (K22). In flame AAS, the program AA-Quality Control used the time profile of the absorption of metal atoms in the flame to detect failures in real-time analysis (K23). Some ES have also been developed for fully automated unattended analyses. A first-class ES for automatic selection and calibration in ICP-AES was built (K24), and the design of a calibration ES for process analytical spectroscopy was discussed (K25). A couple of papers dealt with the implementation of ES for on-line automation and fault detection of flow injection systems in bioprocess monitoring (K26, K27). Another task that can be assigned to ES is the elucidation of structures from spectra or physicochemical parameters. Using two-dimensional NMR data as input, the automated structure elucidation system CHEMICS was able to generate threedimensional structures, therefore allowing the chemist to have an idea of the rough conformation of the candidate structures (K28). The ES CISOC-SES exploited two-dimensional NMR spectra and long-range distance constraints for the elucidation of compounds containing up to 50 nonhydrogen atoms (K29). Three-dimensional NMR spectra of proteins combined with qualitative chemical information and constraint propagation methods allowed the ES AUTOASSIGN to determine the sequential order of amino acid spin systems (K30). Qualitative identification of radionuclides and quantitative determination of spectrum components from γ spectra was performed by an ES with a rule base of 50 criteria (K31). EXPIRS was designed for the interpretation of IR spectra (K32) and subsequently modified in order to improve its speed and predictive ability (K33). Moore and coworkers described an ES destined to assist the analyst in infrared sample preparation and identification (K34). Improvements of ESSESA, another ES for structure elucidation from spectra, were reported (K35-K37). An ES was developed for automated analysis of mixtures GC data. This ES used data-processing techniques such as PCR to produce its responses (K38). Toxic organic compounds could be identified from low-resolution mass spectra with an ES consisting of a classifier followed by molecular weight estimators, filters, and identification modules. Risks of misclassification were reduced thanks to a library of allowed molecular weights and base peaks (K39). The concept of ES was also employed for the real-time detection of hazardous elements in sand and soils from laser plasma and time-resolved spectroscopic data (K40). ESESOC and ESESOC-II were designed for the elucidation of organic compound structures. From input data such as molecular formula or substructure constraints, the ES produced exhaustive and nonredundant lists of candidate structures (K41, K42). Other various applications of ES in chemistry were reported in the past two years. For example a French group used an ES for the automation of open vessel focused microwave digestion. The database guidelines were derived from statistical analysis of 800 procedures (K43). The automatic system HORACE was built for hierarchical classification of chemical reactions based on topological and physicochemical parameters (K44). The consequences of chemical spill accidents can rapidly become disastrous when human decisions have to be made under stress and time constraints. This is a typical situation where an ES can be an invaluable asset to produce a fast and efficient response. In this perspective, Zhu and Stillman developed ERexpert, an ES based on factual information and heuristic knowledge (K45, K46).
ANNs are also popular, and their use in chemistry has been increasing exponentially during the last few years. It is interesting to observe that the word “neural network” did not appear in the 1990 Chemometrics review, whereas in the present issue several tens of reported applications of NN can be found, disseminated among all sections. Due to the variety of topologies and learning rules that can be used with ANN, their range of application is extremely large and encompasses most of the situations where large data sets are at the disposal of the chemist. Difficulties in training the network are related to the size of data sets and the ANN topology. The risks of overfitting the training data and overtraining are addressed by Tetko and co-workers (K47). An overview of applications of ANN in chemistry was published (K48), and in a subsequent paper, the Kohonen self-organizing feature map and the Hopfield ANN were discussed in more detail (K49). Burden showed some practical limitations to the use of ANN as model-free mapping devices (K50), and the concept of large-scale holographic ANN for multispectral sensor fusion and high-speed signal processing was presented (K51). Avoiding local minima is a challenge for all users of back-propagation feed-forward ANN, and a procedure called the “Flashcard Algorithm” was developed to remedy this problem, by overrepresenting difficult examples. The authors also focused on the prediction of untrainable conditions in order to modify the NN topology (K52). Blank and Brown presented an adaptive variant of the global, extended Kalman filter as a second-order weight optimization method. Thanks to its adaptive feature, the algorithm was able to detect redundant information, therefore reducing the training time and significantly enhancing the convergence of the ANN (K53). Harrington proposed a temperature-constrained back-propagation ANN that was claimed to be resistant to overfitting and overtraining (K54). An interpretable ANN solution was found by training the network with an algorithm that was a combination of a modified simplex and PLS regression. The examination of PLS loadings, scores, and residuals made outlier detection possible (K55). As an illustration of the diversity of possible applications of ANN, we can mention their reported use for the position calibration of a novel heavy ion detector system (K56), the recognition of unknown compounds from fluorescence data patterns (K57), and as pattern recognition tools for the development of formulation ES in pharmaceutical industry (K58). The notion of fuzzy logic seems particularly suited to an applied science like analytical chemistry, where physicochemical interactions and response ambiguity often render the concept of crisp decision inappropriate. Bangov et al. applied the membership function approach to extract information from 13C NMR spectra. They discussed the combinatorial problems resulting from this method (K59). Otto also investigated the treatment of spectral data using fuzzy logic and illustrated his paper with the functional group analysis of IR spectra (K60). Fuzzy logic was used for the analysis of the concentration of specific gases in different atmospheres from the measurements of an array of semiconductor gas sensors (K61). Fuzzy logic-based rules are sometimes integrated in the knowledge base of ES to create fuzzy ES. Such a fuzzy ES was devised for the automated interpretation of wavelength-dispersive X-ray fluorescence spectra (K62) and for the identification of minerals from X-ray diffractograms (K63). Other AI-related articles dealt with laboratory automation. Lindsey and Corkan designed an automated chemical workstation including a scheduler for the initiation and monitoring of parallel
experiments. The scheduler allowed evaluation of the performance of various hardware modules so that they could be enhanced (K64). Parallel sample test configuration processing in different laboratories was permitted by the development of a robot system supervised by the software package CLARA (K65). Robots and flow injection analysis systems were coupled to perform the fully automated determination of polyphenols in olive oil (K66). Levy et al. wrote an algorithm for automated twodimensional NMR base plane correction. They used a modified two-dimensional perpendicular drop method combined with statistical and logical decision making for the detection of peaks (K67). A robot was designed for the optimization of liquid/liquid extraction of tricyclic amines, using mixture experimental designs (K68). During the review period, some authors carried out comparisons or combinations of several AI techniques previously mentioned. Mulholland et al. compared the performance of two induction ES and a self-organizing ANN for the optimal choice of a detector in ion chromatography. The three methods were reported to give good results (K69). The classification of products of polyethylene cracking by their degree of unsaturation from GC data was carried out by three ES, a knowledge-based ES, a univariate rule-building ES, and a fuzzy multivariate ES. The latter gave the best results (K70). An integrated intelligent instrument based on supercritical fluid technology was designed for the characterization and remediation of contaminated soils. The data obtained were analyzed by ES and ANN (K71). The concept of hybrid modeling of biochemical production processes was presented, based on a set of dynamic differential equations, an ANN, and a fuzzy ES (K72). Finally, Walczak et al. reported the combination of a fuzzy ES and ANN for the interpretation of X-ray fluorescence spectra. On the basis of historic data and relative emission probabilities, the fuzzy ES created a posteriori rules regarding the usefulness of the different lines in a spectrum, and the importance of the fuzzy rules was determined by the feedforward back-propagation ANN (K73).
We are grateful to Mr. Don Stickel of the Chemical Abstracts Service for providing the STN International searches used in preparation of this review. F.D. gratefully acknowledges support from Elf Atochem. Steven D. Brown is Associate Professor of Chemistry at the University of Delaware. In 1978, he received his Ph.D. from the University of Washington. He has served on the faculty at the University of California, Berkeley and Washington State University prior to joining the faculty at the University of Delaware in 1986. He is currently Editor-in-Chief of the Journal of Chemometrics and Editor of the CRC Chemometrics Series. He also serves as an officer in the International Chemometrics Society and as a member of the Chemstatistics committee of the American Statistical Association. His research interests encompass many aspects of the application of computers to chemical analysis, including developing novel methods for quantitative and qualitative analysis using signal processing methods, neural networks and digital filters. Stephen T. Sum received a B.Sc. in Applied Chemistry (mathematics option) from the University of Waterloo in 1990. He then pursued his studies in the computer science department at Carleton University where he earned an M.Sc. in Information and Systems Science in 1993. He is currently studying for the Ph.D. at the University of Delaware. His research interests are in statistical pattern recognition and signal processing.
Analytical Chemistry, Vol. 68, No. 12, June 15, 1996
Frederic Despagne received his engineer degree from the Physics and Chemistry Graduate School of Bordeaux in 1994. The same year he received an Advanced Studies Diploma in Materials Science from the University of Bordeaux. He is currently doing research at the University of Delaware for Elf Aquitaine. His main interests are the development of neural computing techniques in analytical chemistry. In summer 1996 he will join the ChemoAC Ph.D. program at the Free University of Brussels. Barry K. Lavine obtained his B.S. in chemistry at Temple University in Philadelphia in 1976, MS in analytical chemistry at Ohio State University in 1982, and a Ph.D. degree in analytical chemistry from the Pennsylvania State University in 1986. Currently, he is an Associate Professor of Chemistry where he both teaches and conducts research in analytical chemistry. He has published over 50 papers, and his areas of research include pattern recognition, multivariate calibration, chemical sensors, structure activity and structure property relationship studies, and high-performance liquid chromatography. He is on the editorial board of several journals including Chemometrics & Intelligent Instrumentation, and the Journal of Chemistry and Information Science. He is also the assistant editor for chemometrics for Analytical Letters. LITERATURE CITED (A1) Brown, S. D.; Blank, T. B.; Sum, S. T.; Weyer, L. G. Anal. Chem. 1994, 66, 315R-59R. (A2) Blaser, W. W.; Bredeweg, R. A.; Harner, R. S.; LaPack, M. A.; Leugers, A.; Martin, D. P.; Pell, R. J.; Workman, J., Jr.; Wright, L. G. Anal. Chem. 1995, 67, 47-70. (A3) Bourguignon, B.; Sanchez, F. C.; Massart, D. L. Analusis 1993, 21(10), M21-6. (A4) Rao, G. N.; Rao, S. V. V. S.; Babu, A. R.; Rao, R. S. Asian J. Chem. 1992, 4, 99-104. (A5) Franke, R.; Gruska, A. Methods Princ. Med. Chem. 1995, 2, 113-63. (A6) Clementi, S.; Cruciani, S.; Riganelli, D.; Valigi, R. In New Perspectives in Drug Design; Dean, P. M., Jolles, G., Newton, C. G., Eds.; Academic Press: London, 1995; pp 285-310. (A7) Vandeginste, B. G. M. Chemom. Intell. Lab. Syst. 1994, 25, 147-55. (A8) Cheng, J.-K. Bull. Singapore Natl. Inst. Chem. 1994, 22, 8797. (A9) Geladi, P. J. Chemom. 1993, 7, 213-22. (A10) Schlager, K. J.; Ruchti, T. L. Proc. SPIE-Int. Soc. Opt. Eng. 1995, 2386, 208-21. (A11) Kaufmann, P. In Proceedings-Scandinavian Symposium on Lipids 16th; Lambertsen, G., Ed.; Lipidforum: Bergen, Norway, 1991; pp 70-83. (A12) Ashima, T. In Computer Aided Innovation of New Materials 2; Doyama, M., Ed.; North-Holland: Amsterdam, 1993; pp 8959. (A13) Lindberg, N. O.; Lunstedt, T. Drug Dev. Ind. Pharm. 1995, 21, 987-1007. (A14) Wenning, R. J. Trends Anal. Chem. 1994, 13, 446-57. (A15) Brereton, R. G. In Analysis of Contaminants in Edible Aquatic Resources; Kiceniuk, J. W., Ray, S., Eds.; VCH: New York, 1994; pp 29-58. (A16) Vogt, N. B.; Andersen, S.; Shaaning, M.; Vogt, R. D. In Trace Elements in Natural Waters; Salabu, B., Steinnes, E., Eds.; CRC Press: Boca Raton, FL, 1995; pp 71-97. (A17) Sherwood, P. M. A. In Surface Characteristics of Advanced Polymers; Sabbatini, L., Zambonin, P. G., Eds.; VCH: Weinheim, Germany, 1993; pp 257-98. (A18) Hutter, H.; Grasserbauer, M. Chemom. Intell. Lab. Syst. 1994, 24, 99-116. (A19) Strasters, J. K. Chromatogr. Sci. Ser. 1993, 62, 127-61. (A20) McClure, W. F. In Spectroscopic Techniques for Food Analysis; Wilson, R. H., Ed.; VCH: New York, 1994; pp 13-57. (A21) Descales, B.; Cermelli, I.; Llinas, J. R.; Margall, G.; Martens, A. Analusis 1993, 21(9), M25-8. (A22) Oshima, K.; Oka, K.; Pishva, D. In Computer Aided Innovation New Materials 2; Doyama, M., Ed.; North-Holland: Amsterdam, 1993; pp 931-4. (A23) Trebbia, P. Proc. Scott. Univ. Summer Sch. Phys. 1992 (Quantit. Microbeam Anal.) 1993, 40, 189-201. (A24) Converse, J. G. Process Control Qual. 1993, 5, 131-6. (A25) Miller, C. E. Chemom. Intell. Lab. Syst. 1995, 30, 11-22. (A26) Carey, P. W. Trends Anal. Chem. 1994, 13, 210-8. (A27) Henrion, R. Chemom. Intell. Lab. Syst. 1994, 25, 1-23. (A28) Kraus, G.; Goeppert, J.; Rosenstiel, W.; Gauglitz, G. GIT Fachz. Lab. 1994, 38, 1133-8. (A29) Bishop, S. R. Anal. Proc. 1993, 30, 310-4. (A30) Booksh, K. S.; Kowalski, B. R. Anal. Chem. 1994, 66, 782A94A. (A31) Wold, S. Chemom. Intell. Lab. Syst. 1995, 30, 109-16. (A32) Brown, S. D. Chemom. Intell. Lab. Syst. 1995, 30, 49-58. Software (A33) Gonzales-Arjona, D.; Mejias, J. A.; Gonzalez, A. G. Anal. Chim. Acta 1994, 295, 119-25. (A34) Gonzales-Arjona, D.; Mejias, J. A.; Gonzalez, A. G. Anal. Chim. Acta 1994, 297, 473. (A35) Lohninger, H. Chemom. Intell. Lab. Syst. 1994, 22, 147-53. 50R
Analytical Chemistry, Vol. 68, No. 12, June 15, 1996
(A36) Katsumi, H.; Yoshida, M.; Kikuzono, Y.; Takayama, C.; Marsili, M. In Comput. Aided Innovation New Mater.; Doyama, M, Ed.; North-Holland: Amsterdam, 1993; pp 887-90. (A37) Makanyan, O. G.; Karabunerliev, S. H.; Ivanov, J. M.; Dimitrov, O. N. Comput. Chem. 1994, 18, 173-87. (A38) Varmuza, K.; Scsibrany, H. Lab. Microcomput. 1993, 12, 99103. (A39) Gort, S. M.; Hoogerbrugge, R. Chemom. Intell. Lab. Syst. 1995, 28, 193-9. (A40) Rohrer, C.; Wegscheider, W. GIT Fachz. Lab. 1994, 38, 68891. (A41) Walsh, S.; Diamond, D. Talanta 1995, 42, 561-72. (A42) Majoras, R. E.; Richardson, W. M.; Semour, R. S. J. Radioanal. Nucl. Chem. 1995, 193, 207-10. (A43) Bright, D. S. Microbeam Anal. 1995, 4, 151-63. (A44) Rasband, W. S.; Bright, D. S. Microbeam Anal. 1995, 4, 13749. (A45) Lowney, J. R. Microbeam Anal. 1995, 4, 131-6. (A46) Wise, B. M. Process Control Qual. 1993, 5, 73-85. (A47) Sarabia, L.; Ortiz, M. C. Trends Anal. Chem. 1994, 13, 1-6. (A48) Phillips, G. F. Anal. Proc. 1994, 31, 222. Books (A49) Frank, I. E.; Todeschini, R. The Data Analysis Handbook; Elsevier: Amsterdam, 1994. (A50) Meloun, M., Militky, J., Forina, M., Eds. Chemometrics for Analytical Chemistry; Horwood: London, 1994; Vol. 2. (A51) Buydens, L. M. C., Melssen, W. J., Eds. Chemometrics: Exploring and Exploiting Chemical Information; Catholic University Press: Nijmegen, 1994. (A52) Collier, H., Ed. Further Advances in Chemical Information; Royal Society of Chemistry: Cambridge, UK, 1994. (A53) van de Waterbeemd, H., Ed. Chemometric Methods in Molecular Design; VCH: Weinheim, Germany, 1995. (A54) Adams, M. J. Chemometrics in Analytical Spectroscopy; Royal Society of Chemistry: Cambridge, UK, 1995. (A55) George, W. O.; Steele, D. Computing Applications in Molecular Spectroscopy, 2nd ed.; Royal Society of Chemistry: Cambridge, UK, 1995. (A56) Wilkins, C. L., Ed. Computer-Enhanced Analytical Spectroscopy; Plenum Press: New York, 1994; Vol. 4. (A57) Bellomo, N.; Preziosi, L. Modelling, Mathematical Methods and Scientific Computation; CRC Press: Boca Raton, FL, 1994. (A58) Sigmon, K. MATLAB Primer, 4th ed.; CRC Press: Boca Raton, FL, 1994. (A59) Marazzi, A. Algorithms, Routines and S Functions for Robust Statistics; Chapman and Hall: New York, 1994. (A60) Everitt, B. A Handbook of Statistical Analyses Using S-Plus; Chapman and Hall: New York, 1994. (A61) Venables, W. N.; Ripley, B. D. Modern Applied Statistics with S-Plus; Springer Verlag: New York, 1995. (A62) Ha¨rdle, W.; Klinke, S.; Turlach, B. A. XploRe: An Interactive Statistical Computing Environment; Springer Verlag: New York, 1995. (A63) Korsan, R. J. Decision Support Systems with Mathematica; TelossSpringer Verlag: New York, 1994. (A64) Lau, H. T. A Numerical Library in C for Scientists and Engineers; CRC Press: Boca Raton, FL, 1994. (A65) Walter, G. G. Wavelets and Other Orthogonal Systems with Applications; CRC Press: Boca Raton, FL, 1994. (A66) Cleveland, W. S. The Elements of Graphing Data; Chapman and Hall: New York, 1994. (A67) Schiff, D.; D’Agostino, R. Practical Engineering Statistics; John Wiley: New York, 1995. (A68) Hinkelmann, K.; Kempthorne, O. Design and Analysis of Experiments; John Wiley and Sons: New York, 1994. (A69) Bissell, D. Statistical Methods for SPC and TQM; Chapman and Hall: New York, 1994. (A70) Mittag, H.-J.; Rinne, H. Statistical Methods of Quality Assurance; Chapman and Hall: New York, 1994. (A71) Mandel, J. The Analysis of Two-Way Layouts; Chapman and Hall: New York, 1994. (A72) Mesterson-Gibbons, M. A Concrete Approach to Mathemical Modeling, 2nd ed.; John Wiley and Sons: New York, 1995. (A73) Birkes, D.; Dodge, Y. Alternative Methods of Regression; John Wiley: New York, 1993. (A74) Rencher, A. C. Methods of Multivariate Analysis; John Wiley and Sons: New York, 1995. (A75) Everitt, B. S.; Dunn, G. Applied Multivariate Data Analysis; John Wiley and Sons: New York, 1995. (A76) Barnett, V.; Lewis, T. Outliers in Statistical Data; John Wiley and Sons: New York, 1994. (A77) Chong, E. K. P.; Zak, S. H. An Introduction to Optimization; John Wiley and Sons: New York, 1995. (A78) Myers, R. H.; Montgomery, D. C. Response Surface Methodology: Process and Product Optimization Using Designed Experiments; John Wiley and Sons: New York, 1995. (A79) Shao, J.; Tu, D. The Jackknife and the Bootstrap. Springer Verlag: New York, 1995. (A80) Hjorth, J. S. U. Computer Intensive Statistical Methods: Validation, Model Selection and Bootstrap; Chapman and Hall: New York, 1994. (A81) Huberty, C. J. Applied Discriminant Analysis; John Wiley and Sons: New York, 1994.
(A82) Weizman, E. A. Computer Programs for Qualitative Data Analysis; Sage Publications: Thousand Oaks, CA, 1995. (A83) Fuller, W. A. Introduction to Statistical Time Series, 2nd ed.; John Wiley: New York, 1995. (A84) Hand, D. J., Ed. Artificial Intelligence Frontiers in Statistics; Chapman and Hall: New York, 1994. (A85) Bulsari, A. B., Ed. Neural Networks for Chemical Engineers; Elsevier: New York, 1995. (A86) Pham, D. T.; Xing, L. Neural Networks for Prediction and Control; Springer Verlag: London, 1995. (A87) Bishop, C. Neural Networks for Pattern Recognition; Oxford University Press: Oxford, UK, 1995. (A88) Masters, T. Advanced Algorithms for Neural Networks: A C++ Sourcebook; John Wiley and Sons: New York, 1995. (A89) Weistead, S. T. Neural Network and Fuzzy Logic Applications in C/C++; John Wiley and Sons: New York, 1994. (A90) Pandya, A. S.; Macy, R. B. Pattern Recognition with Neural Networks in C++; CRC: Boca Raton, FL, 1995. (A91) Cooper, J. W.; Lam, R. B. A Jump Start Course in C++ Programming; John Wiley and Sons: New York, 1994. (A92) Gilster, P. The New Internet Navigator; John Wiley and Sons: New York, 1995. (A93) Massart, D. L.; Lewi, P. J. Principal Components: An Audiovisual Course; Elsevier: Amsterdam, 1994. Tutorials (A94) Mark, H.; Workman, J., Jr. Spectroscopy 1994, 9(5), 22-3. (A95) Mark, H.; Workman, J., Jr. Spectroscopy 1994, 9(8), 26-7. (A96) Meloun, M.; Militky, J. Chem. Pap. 1994, 48, 164-9; Chem. Abstr. 1995, 122, 229417. (A97) Chau, F.; Chung, W. H. J. Chem. Educ. 1995, 72, A84-5. (A98) Mitra, S.; Bose, T. Chemom. Intell. Lab. Syst. 1994, 22, 3-16. (A99) Brereton, R. G. Analyst 1994, 119, 2149-60. (A100) Kragten, J. Analyst 1994, 119, 2161-6. (A101) Brereton, R. G. Analyst 1995, 120, 2313-36. (A102) Kourti, T.; MacGregor, J. F. Chemom. Intell. Lab. Syst. 1995, 28, 3-22. (A103) Thomas, E. V. Anal. Chem. 1994, 66, 795A-804A. (A104) Brown, S. D. Appl. Spectrosc. 1995, 49(12), 14A-31A. (A105) Ho ¨skuldsson, A. Chemom. Intell. Lab. Syst. 1994, 23, 1-28. (A106) Smits, J. R. M.; Melssen, W. J.; Buydens, L. M. C.; Kateman, G. Chemom. Intell. Lab. Syst. 1994, 23, 165-90. (A107) Melssen, W. J.; Smits, J. R. M.; Buydens, L. M. C.; Kateman, G. Chemom. Intell. Lab. Syst. 1994, 23, 267-92. STATISTICS (B1) Ouchi, G. I. LC-GC 1994, 12, 826-30. (B2) Howarth, R. J. Analyst 1995, 120, 1851-73. (B3) Maier, E. A.; Quevauviller, Ph.; Griepink, B. Anal. Chim. Acta 1993, 283, 590-9. (B4) Meloun, M.; Militky, J. Chem. Pap. 1994, 48, 151-7. (B5) Meloun, M.; Militky, J. Chem. Pap. 1994, 48, 158-63. (B6) Meloun, M.; Militky, J. Chem. Pap. 1994, 48, 164-9. (B7) Meloun, M.; Militky, J. Chem. Pap. 1994, 49, 68-75. (B8) Danzer, K. In Akkreditierung und Qualitaetssicherung in der Analytischen Chemie; Guenzler, H., Ed.; Springer: Berlin, 1994; pp 71-103; Chem. Abstr. 1995, 123, 101396. (B9) Bunzl, K. Low-Level Meas. Radioact. Environ., Proc. Int. Summer Sch., 3rd, Meeting Date 1993; Garcia-Leon, M., Garcia-Tenorio, R., Eds; World Sci.: Singapore, 1994; pp 389415. (B10) Wegscheider, W. In Akkreditierung und Qualitaetssicherung in der Analytischen Chemie; Guenzler, H., Ed; Springer: Berlin, 1994; pp 105-29; Chem. Abstr. 1995, 123, 101397. (B11) Grimstvedt, A. M. Bull.-Nor. Geol. Unders. 1995, 427, 96-9; Chem. Abstr. 1995, 123, 245296. (B12) Rymen, T. Tech. Instrum. Anal. Chem. 1995, 17, 591-639. (B13) Toelg, G.; Tschoepel, P. In Determination of Trace Elements; Alfassi, Z. B., Ed; VCH: Weinheim, Germany, 1994; pp 1-38. (B14) Analytical Methods Committee, Royal Society of Chemistry. Analyst 1995, 120, 2303-8. (B15) Xie, Y.; Wang, J.; Liang, Y.; Yu, R. Fenxi Huaxue 1994, 22, 294-300; Chem. Abstr. 1995, 121, 98438. (B16) Grize, Y. L. J. Chemom. 1995, 9, 239-62. (B17) Heydorn, K. Anal. Chim. Acta 1993, 283, 494-9. (B18) Mudd, J. L.; Baechtel, F. S.; Duewer, D. L.; Currie, L. A.; Reeder, D. J.; Leigh, S. D.; Liu, H. K. Anal. Chem. 1994, 66, 3303-17. (B19) Duewer, D. L.; Currie, L. A.; Reeder, D. J.; Leigh, S. D.; Liu, H.-K.; Mudd, J. L. Anal. Chem. 1995, 67, 1220-31. (B20) Cofino, W. P.; Wells, D. E. Mar. Pollut. Bull. 1994, 29, 14958. (B21) Geilenkeuser, W. J.; Roehle, G. Eur. J. Clin. Chem. Clin. Biochem. 1994, 32, 369-75. (B22) Dvorkin, V. I. Chemom. Intell. Lab. Syst., 1994, 22, 127-46. (B23) Feinberg, M. Analusis 1994, 22(5), M34-7. (B24) Feinberg, M.; Bugner, E.; Theiller, G.; Houba, V.; Kadjik, F. J. Chemom. 1995, 9, 197-209. (B25) Gaskin, J. E. Analyst 1994, 119, 1531-5. (B26) Minkkinen, P. Chemom. Intell. Lab. Syst. 1995, 29, 263-70. (B27) IUPAC, Anal., Appl. Clin. Chem. Div. Pure Appl. Chem. 1995, 67, 331-43. (B28) Massart, D. L.; Smeyers-Verbeke, J.; Vandeginste, B. Analusis 1994, 22(5), M14-6.
(B29) Rodriguez, L. C.; Campana, A. M. C.; Barrero, F. A.; Linares, C. J.; Ceba, M. R. J. AOAC Int. 1995, 78, 471-6. (B30) Grize, Y. L.; Schmidli, H.; Born, J. J. Chromatogr., A 1994, 686, 1-10. (B31) Gonzalez, A. G.; Asuero, A. G. Fresenius’ J. Anal. Chem. 1993, 346, 885-7. (B32) Ellison, S. L. R.; Cox, M. G.; Forbes, A. B.; Butler, B. P.; Hannaby, S. A.; Harris, P. M.; Hodson, S. M. J. AOAC Int. 1994, 77, 777-81. (B33) Hartmann, C.; Massart, D. L.; McDowall, R. D. J. Pharm. Biomed. Anal. 1994, 12, 1337-43. (B34) Lang, J. R.; Bolton, S. M. Prog. Pharm. Biomed. Anal. 1994, 1, 345-67. (B35) Dadgar, D.; Burnett, P. E.; Choc, M. G.; Gallicano, K.; Hooper, J. W. J. Pharm. Biomed. Anal. 1995, 13, 89-97. (B36) Kuselman, I.; Shenhar, A. Anal. Chim. Acta 1995, 306, 3015. (B37) Bauer, K. H. Vom Wasser 1994, 82, 393-408; Chem. Abstr. 1995, 122, 121999. (B38) Eaton, A.; Sanders, P. M. Proc.-Water Quality Technology Conference (Part 1); 1993; pp 731-44. (B39) Hinton, S. W. Tappi J. 1994, 77, 83-90; Chem. Abstr. 1995, 121, 91155. (B40) Ferru´s, R.;Egea, M. R. Anal. Chim. Acta 1994, 287, 119-45. (B41) Gautsch, K.; Keller, B.; Keller, H.; Pei, P.; Vonderschmitt, D. J. Eur. J. Clin. Chem. Clin. Biochem. 1993, 31, 433-40. (B42) Chung, C. J. F. Math. Geol. 1993, 25, 851-65. (B43) Ramsey, M. H.; Argyraki, A.; Thompson, M. Analyst 1995, 120, 2309-12. (B44) Argyraki, A.; Ramsey, M. H.; Thompson, M. Analyst 1995, 120, 2799-803. (B45) Thompson, M.; Ramsey, M. H. Analyst 1995, 120, 261-70. (B46) Ramsey, M. H. In Environmental Sampling for Trace Analysis; Market, B., Ed; VCH: Weinheim, Germany, 1994; pp 93-108. (B47) Ramsey, M. H.; Thompson, M. Analyst 1995, 120, 1353-6. (B48) Van Der Meulen, E. A.; Boogaard, P. J.; Van Sittert, N. J. Clin. Chem. 1994, 40, 1698-702. (B49) Scott, E. M. In Low-Level Meas. Radioact. Environ. Proc., Int. Summer Sch., 3rd, Meeting Date 1993; Garcia-Leon, M., Garcia-Tenorio, R., Eds.; World Sci.: Singapore, 1994; pp 41729. (B50) Chen, J.; Hopmans, J. W.; Fogg, G. E. Soil Sci. 1995, 159, 155-61. (B51) Colin, F.; Jauzein, M.; Grapin, G. Quim. Anal. (Barcelona) 1994, 13, S80-8; Chem. Abstr. 1995, 122, 221937. (B52) Szucs, A.; Jordan, G. Water Sci. Technol. 1994, 30, 73-8. (B53) Somerville, M. C.; Evans, E. G. Atmos. Environ. 1995, 29, 2429-38. (B54) Doerffel, K. Fresenius’ J. Anal. Chem. 1994, 348, 183-7. (B55) Johnson, R. L.; Latimer, G. W., Jr.; Spiegelman, C. J. AOAC Int. 1994, 77, 1660-3. (B56) Hueck, M. Materialwiss. Werkstofftech. 1994, 25, 20-9; Chem. Abstr. 1995, 120, 207411. (B57) Kragten, J. Chemom. Intell. Lab. Syst. 1995, 28, 89-97. (B58) Thompson, M.; Lowthian, P. J. Analyst 1995, 120, 271-2. (B59) Straume, M.; Veldhuis, J. D.; Johnson, M. L. Methods Enzymol. 1994, 24, 121-50. (B60) Cammann, K. Sens. Actuators, B 1995, B25, 769-72. (B61) Meloun, M.; Militky, J. Anal. Chim. Acta 1994, 293, 183-9. (B62) Wa¨tzig, H. J. Chromatogr., A 1995, 700, 1-7. (B63) Baumann, K.; Watzig, H. J. Chromatogr., A 1995, 700, 9-20. (B64) Baeza-Baeza, J. J.; Ramis-Ramos, G. Anal. Chim. Acta 1995, 316, 173-84. (B65) Oostrom, M.; Dane, J. H.; Missildine, B. C.; Lenhard, R. J. Soil Sci. 1995, 160, 28-42. (B66) Hardy, A. J.; Wegscheider, W.; Haswell, S. J.; Hailey, P. A. Analyst 1995, 120, 1875-84. (B67) Cabral, J. P. S. Analusis 1994, 22(1), 1-4. (B68) Ignacio Garcia Alonso, J. Anal. Chim. Acta 1995, 312, 5778. (B69) Hayashi, Y.; Matsuda, R. Anal. Sci. 1994, 10, 881-8. (B70) Svergun, D. I.; Pedersen, J. S. J. Appl. Crystallogr. 1994, 27, 241-8. (B71) Kvalheim, O. M.; Brakstad, F.; Liang, Y. Anal. Chem. 1994, 66, 43-51. (B72) Rietjens, M. Anal. Chim. Acta 1995, 316, 205-15. (B73) Faber, N. M.; Buydens, L. M. C.; Kateman, G. J. Chemom. 1993, 7, 495-526. (B74) Booksh, K.; Kowalski, B. R. J. Chemom. 1994, 8, 45-63. (B75) Faber, N. M.; Buydens, L. M. C.; Kateman, G. J. Chemom. 1994, 8, 181-203. (B76) Faber, N. M.; Buydens, L. M. C.; Kateman, G. J. Chemom. 1994, 8, 373. (B77) Faber, N. M.; Meinders. M. J.; Geladi, P.; Sjo ¨stro¨m, M.; Buydens, L. M. C.; Kateman, G. Anal. Chim. Acta 1995, 304, 257-71. (B78) Faber, N. M.; Meinders. M. J.; Geladi, P.; Sjo ¨stro¨m, M.; Buydens, L. M. C.; Kateman, G. Anal. Chim. Acta 1995, 304, 273-83. (B79) De Vries, S.; Ter Braak, C. J. F. Chemom. Intell. Lab. Syst. 1995, 30, 239-45. (B80) Kalantar, A. H.; Gelb, R. I.; Alper, J. S. Talanta 1995, 42, 597603. (B81) Riu, J.; Rius, F. X. J. Chemom. 1995, 9, 343-62. (B82) Mernet, J.-M. Spectrochim. Acta, Part B 1994, 49B, 131324.
Analytical Chemistry, Vol. 68, No. 12, June 15, 1996
(B83) Chambless, D. A.; Dubose, S. S.; Sensintaffar, E. L. Health Phys. 1994, 66, 313-7. (B84) Currie, L. A. Nucl. Instrum. Methods Phys. Res., Sect. B 1994, 92, 188-93. (B85) Gatz, D. F.; Smith, L. Atmos. Environ. 1995, 29, 1185-93. (B86) Gatz, D. F.; Smith, L. Atmos. Environ. 1995, 29, 1195-200. (B87) Veltkamp, D. J. Process Control Qual. 1993, 5, 205-17. (B88) Puebla, C. Chemom. Intell. Lab. Syst. 1994, 26, 27-35. (B89) Sebastian, P. R.; Booth, D. E.; Hu, M. Y. J. Chem. Inf. Comput. Sci. 1994, 34, 881-9. (B90) Sebastian, P. R.; Booth, D. E.; Hu, M. Y. J. Chem. Inf. Comput. Sci. 1995, 35, 442-50. (B91) Nomikos, P.; MacGregor, J. F. Chemom. Intell. Lab. Syst. 1995, 30, 97-108. (B92) Ku, W.; Storer, R. H.; Georgakis, C. Chemom. Intell. Lab. Syst. 1995, 30, 179-96. (B93) Raich, A. C.; C¸ inar, A. Chemom. Intell. Lab. Syst. 1995, 30, 37-48. (B94) Staats, G. Fresenius’ J. Anal. Chem. 1995, 352, 413-9. (B95) Chakravarty, T. Chem. Eng. Prog. 1993, 89, 76-83. (B96) Liebich, V.; Kucharkowski, R. Chemom. Intell. Lab. Syst. 1994, 24, 197-203. (B97) Vander Heyden, Y.; Luypaert, K.; Hartmann, C.; Massart, D. L.; Hoogmartens, J.; De Beer, J. Anal. Chim. Acta 1995, 312, 245-62. (B98) Erickson, W. P.; McDonald, L. L. Environ. Toxicol. Chem. 1995, 14, 1247-56. (B99) Dhaliwal, B. S.; Dolan, R. J.; Smith, R. W. Water Environ. Res. 1995, 67, 953-63. (B100) Vandeginste, B. G.; Quadt, J. F. A. Analusis 1994, 22(5), M30-3. (B101) Analytical Methods Committee, Royal Society of Chemistry. Analyst 1995, 120, 29-34. (B102) Leardi, R. J. Chemom. 1994, 8, 65-79. (B103) Walczak, B.; Massart, D. L. Chemom. Intell. Lab. Syst. 1995, 27, 41-54. (B104) Walczak, B. Chemom. Intell. Lab. Syst. 1995, 28, 259-72. (B105) Walczak, B. Chemom. Intell. Lab. Syst. 1995, 29, 63-73. OPTIMIZATION (C1) (C2) (C3) (C4) (C5) (C6) (C7) (C8) (C9) (C10) (C11) (C12) (C13) (C14) (C15) (C16) (C17) (C18) (C19) (C20) (C21) (C22) (C23) (C24) (C25) (C26) (C27) (C28) (C29) (C30) (C31) (C32)
Atkinson, A. C. Chemom. Intell. Lab. Syst. 1995, 28, 35-47. Stone, R. A.; Veevers, A. J. Chemom. 1994, 8, 103-10. Rozycki, C. Chem. Anal. 1993, 38, 681-98. Curtis, M. A. J. Chem. Educ. 1994, 71, 775-8. De Boer, J. H. Pharm. World Sci. 1993, 15, 180-3. Gennaro, M. C. Adv. Chromatogr. 1995, 35, 343-81. Hayashi, Y.; Matsuda, R. Adv. Chromatogr. 1994, 34, 347427. Brochon, J.-C. Methods Enzymol. 1994, 240, 262-311. Lucasius, C. B.; Kateman, G. Chemom. Intell. Lab. Syst. 1994, 25, 99-145. de Vanssay, E; Zubrzycki, S.; Sternberg, R.; Raulin, F.; Sergent, M.; Phan-Tan-Luu, R. J. Chromatogr., A 1994, 688, 161-70. Nechar, M.; Molina, M. F. M.; Rodriguez, L. C.; Sendra, J. M. B. Anal. Chim. Acta 1995, 316, 185-93. Wieling, J.; Jonkman, J. H. G.; Hempenius, J.; Mensink, C. K. Chemom. Intell. Lab. Syst. 1994, 25, 355-66. Vojnovic, D.; Campisi, B.; Mattei, A.; Favretto, L. Chemom. Intell. Lab. Syst. 1995, 27, 205-10. Papamichael, E. M.; Evmiridis, N. P.; Thanasoulias, N.; Stefanou, D. Chemom. Intell. Lab. Syst. 1995, 30, 227-37. Kaufmann, P.; Kowalski, B. R.; Alander, J. Chemom. Intell. Lab. Syst. 1994, 23, 331-9. Kaufmann, P. Chemom. Intell. Lab. Syst. 1995, 27, 105-14. Guillaume, Y.; Guinchard, C. J. Chromatogr. Sci. 1995, 33, 204-10. Nagy-Turak, A.; Vegh, Z. J. Chromatogr., A 1994, 688, 5017. Michaud, J. P.; Gandolfi, A. J.; Brendel, K. Life Sci. 1994, 55, 635-51. Michaud, J. P.; Gandolfi, A. J.; Brendel, K. Toxicol. Methods 1995, 5, 21-40. Schlosser, P. J. Toxicol. Environ. Health 1994, 43, 495-530. Rasemann, W.; Seltmann, U.; Hempel, M. Fresenius’ J. Anal. Chem. 1995, 351, 632-41. Gustavo Gonzalez, A.; Gonzalez-Arjona, D. Anal. Chim. Acta 1994, 298, 65-73. Langsrud, O.; Ellekjaer, M. R.; Næs, T. J. Chemom. 1994, 8, 205-19. Rogan, M. M.; Altria, K. D.; Goodall, D. M. Chromatographia 1994, 38, 723-9. Arnoldsson, K. C.; Kaufmann, P. Chromatographia 1994, 38, 317-24. Vander Heyden, Y.; Hartmann, C.; Massart, D. L.; Michel, L.; Kiechle, P.; Erni, F. Anal. Chim. Acta 1995, 316, 15-26. Oles, P. J. J. AOAC Int. 1993, 76, 615-20. Chen, C.-Y.; Ling, Y.-C. J. High Resolut. Chromatogr. 1994, 17, 784-91. Vega, M.; Pardo, R.; Barrado, E.; de la Fuente, M. A.; del Valle, J. L. Fresenius’ J. Anal. Chem. 1994, 350, 139-44. Lan, W. G.; Wong, M. K.; Chen, N.; Sin, Y. M. Analyst 1994, 119, 1659-67. Lan, W. G.; Wong, M. K.; Chen, N.; Sin, Y. M. Analyst 1994, 119, 1669-75.
Analytical Chemistry, Vol. 68, No. 12, June 15, 1996
(C33) Lan, W. G.; Wong, M. K.; Chen, N.; Sin, Y. M. Analyst 1995, 120, 115-24. (C34) Lan, W. G.; Wong, M. K.; Chee, K. K.; Sin, Y. M. Analyst 1995, 120, 273-9. (C35) Lan, W. G.; Wong, M. K.; Chen, N.; Sin, Y. M. Talanta 1994, 41, 1917-27. (C36) Lan, W. G.; Chee, K. K.; Wong, M. K.; Lee, H. K.; Sin, Y. M. Analyst 1995, 120, 281-7. (C37) Chee, K. K.; Lan, W. G.; Wong, M. K.; Lee, H. K. Anal. Chim. Acta 1995, 312, 271-80. (C38) Zhou, C. Y.; Wong, M. K.; Koh, L. L.; Wee, Y. C. Anal. Chim. Acta 1995, 314, 121-30. (C39) Wan, H. B.; Wong, M. K.; Mok, C. Y. J. Chromatogr. Sci. 1995, 33, 66-70. (C40) Galley, P. J.; Horner, J. A.; Hieftje, G. M. Spectrochim. Acta, Part B 1995, 50B, 87-107. (C41) van der Valde-Koerts, T.; de Boer, J. L. Anal. At. Spectrom. 1994, 9, 1093-8. (C42) Ford, M. J.; Ebdon, L.; Hutton, R. C.; Hill, S. J. Anal. Chim. Acta 1994, 285, 23-31. (C43) Pergantis, S. A.; Cullen, W. R.; Wade, A. P. Talanta 1994, 41, 205-9. (C44) Vera Tome, F.; Jurado Vargas, M.; Martin Sanchez, A. Nucl. Instrum. Methods Phys. Res., Sect. A 1994, 348, 183-7. (C45) San Vicente, A.; Arranz, A.; Moreda, J. M.; Arranz, J. F. Anal. Chim. Acta 1994, 298, 87-90. (C46) Gonzalez-Arjona, D.; Gustavo, Gonzalez, A. Anal. Chim. Acta 1994, 293, 205-10. (C47) Lopez Molinero, A.; Rubio, C.; Castillo, J. R. Anal. Chim. Acta 1994, 297, 417-25. (C48) Miyamae, Y.; Zhang, C.; Narusawa, Y. J. Chem. Software 1995, 2, 159-66; Chem. Abstr. 1995, 123, 20680. (C49) Andersson, R.; Haemaelaeinen, M. D. Chemom. Intell. Lab. Syst. 1994, 22, 49-61. (C50) Woodruff, T. J.; Bois, F. Y. Toxicol. Lettt. 1993, 69, 181-96. (C51) Utteridge, T. D.; Ralston, A. C.; Wallace, S. A.; Allen, B. J.; Beddoe, A. H. Appl. Radiat. Isot. 1994, 45, 789-94. (C52) Boerjesson, J.; Jonson, R.; Mattsson, S.; Muentzing, K.; Toelli, H. Basic Life Sci. 1993, 60, 267-73. (C53) Bernasconi, G.; Bamford, S. A.; Dosan, B.; Haselberger, N.; Markowicz, A.; Mahmoud, A.; Valkovic, V. X-Ray Spectrom. 1994, 23, 65-70. (C54) Beran, P.; Szoeke, A. Acta Crystallogr., Sect. A: Fundam. Crystallogr. 1995, A51, 20-7. (C55) Harris, K. D. M.; Tremayne, M.; Lightfoot, P.; Bruce, P. G. J. Am. Chem. Soc. 1994, 116, 3543-7. (C56) Rous, P. J. Surf. Sci. 1993, 296, 358-73. (C57) Ulyanov, N.; Nikolai, B.; Schmitz, U.; James, T. L. J. Biomol. NMR 1993, 3, 547-68. (C58) Kraulis, P. J. J. Mol. Biol. 1994, 243, 696-718. (C59) Morelle, N.; Brutscher, B.; Simorre, J.-P.; Marion, D. J. Biomol. NMR 1995, 5, 154-60. (C60) Christakos, G.; Killam, B. R. Water Resour. Res. 1993, 29, 4063-76. (C61) Zellers, E. T.; Batterman, S. A.; Han, M.; Patrash, S. J. Anal. Chem. 1995, 67, 1092-106. (C62) Ho ¨rchner, U.; Kalivas, J. H. J. Chemom. 1995, 9, 283-308. (C63) Wang, S.-H.; Xiao, D.; Yu, R. Chem. Res. Chin. Univ. 1993, 9, 296-302. (C64) Xie, Y.-L.; Wang, J.-H.; Liang, Y.-Z.; Ge, K.; Yu, R. J. Chemom. 1993, 7, 369-79. (C65) Xie, Y.-L.; Wang, J.-H.; Liang, Y.-Z.; Sun, L.-X.; Song, X.-H., Yu, R. J. Chemom. 1993, 7, 527-41. (C66) Kvasnicka, V.; Pospı´chal, J. J. Chemom. 1995, 9, 309-22. (C67) Cong, P.; Li, T. Anal. Chim. Acta 1994, 293, 191-203. (C68) Martin, S.; Rivory, J.; Schoenauer, M. Opt. Commun. 1994, 110, 503-6. (C69) Desjarlais, J. R.; Handel, T. M. Protein Sci. 1995, 4, 200618. (C70) Lucasius, C. B.; Dane, A. D.; Kateman, G. Anal. Chim. Acta 1993, 282, 647-69. (C71) Jouan-Rimbaud, D.; Massart, D.-L.; Leardi, R.; De Noord, O. E. Anal. Chem. 1995, 67, 4295-301. (C72) Lucasius, C. B.; Beckers, M. L. M.; Kateman, G. Anal. Chim. Acta 1994, 286, 135-53. (C73) Ho¨rchner, U.; Kalivas, J. H. Anal. Chim. Acta 1995, 311, 1-13. (C74) Andrade, J. M.; Prada, D.; Muniategui, S.; Gomez, B.; Pan, M. J. Chemom. 1993, 7, 427-38. (C75) Costadinnova, L.; Nedeltcheva, T. Analyst 1995, 120, 221720. (C76) Araujo, A.; Lima, J. L. F. C.; Gracia, J.; Poch, M.; Alonso, J.; Bartroli, J.; del Valle, M. Anal. Chim. Acta 1995, 310, 28996. (C77) Duineveld, C. A. A.; Bruins, C. H. P.; Smilde, A. K.; Bolhuis, G. K.; Zuurman, K.; Doornbos, D. A. Chemom. Intell. Lab. Syst. 1994, 25, 183-201. (C78) Duineveld, C. A. A.; Coenegracht, P. M. J. Chemom. Intell. Lab. Syst. 1995, 30, 23-36. (C79) Buice, R. G., Jr.; Pinkston, P.; Lodder, R. A. Appl. Spectrosc. 1994, 48, 517-24. (C80) Olson, J. B., Jr.; Markley, J. L. J. Biomol. NMR 1994, 4, 385410. (C81) Xu, Y.; Sugar, I. P.; Krishna, N. R. . J. Biomol. NMR 1995, 5, 37-48. (C82) Dennoyer, E. R. At. Spectrosc. 1994, 15, 7-16.
(C83) Celentano, F. C.; Gianazza, E. Appl. Theor. Electrophor. 1993, 3, 191-201. (C84) Westrich, B. J.; Buzzard, I. M.; Gatewood, L. C.; McGovern, P. G. J. Food Compos. Anal. 1994, 7, 223-39. (C85) Hudak, P. F. Environ. Monit. Assess. 1994, 30, 197-210. (C86) Suwondo, E.; Floquet, P.; Pibouleau, L.; Domenech, S. Chem. Eng. Commun. 1995, 134, 73-85. (C87) Wang, Q.-S.; Yan, B.-W.; Zhang, Z.-C. J. Planar Chromatogr.Mod. TLC 1994, 7, 229-32. (C88) Wang, Q.-S.; Zhan, Z.-P. J. Planar Chromatogr.-Mod. TLC 1994, 7, 394-8. (C89) Wang, Q.-S.; Yan, B.-W.; Zhang, Z.-C. J. Planar Chromatogr.Mod. TLC 1994, 7, 477-80. (C90) Wang, Q.-S.; Yan, B.-W.; Zhang, L. Chromatographia 1995, 40, 463-6. (C91) Han-Xi, S.; Guo-Sheng, Y.; Ru-Yu, G.; Qin-Sun, W. Chromatographia 1995, 40, 303-6. (C92) Ong, C. P.; Chow, K. K.; Ng, C. L.; Ong, F. M.; Lee, H. K.; Li, S. F. Y. J. Chromatogr., A 1995, 692, 207-12. (C93) Akard, M.; Sacks, R. Anal. Chem. 1995, 67, 2733-7. (C94) Martinez-Vidal, J. L.; Parilla, P.; Fernandez-Alba, A. R.; Carreno, R.; Herrara, F. J. Liq. Chromatogr. 1995, 18, 2969-89. (C95) Sahota, R. S.; Khaledi, M. G. Anal. Chem. 1994, 66, 237481. (C96) Lundstedt, T.; Thelin, B. Chemom. Intell. Lab. Syst. 1995, 29, 255-61. (C97) Kettaneh-Wold, N.; MacGregor, J. F.; Dayal, B.; Wold, S. Chemom. Intell. Lab. Syst. 1994, 23, 39-50. (C98) David, V. Rev. Roum. Chim. 1993, 38, 527-31. (C99) Frederick, B. G.; Richardson, N. V. J. Electron Spectrosc. Relat. Phenom. 1995, 73, 149-55. (C100) Tu, X. M.; Meng, X.-L.; Pogano, M. J. Chemom. 1994, 8, 36570. (C101) Power, J. F.; Prystay, M. C. Appl. Spectrosc. 1995, 49, 70924. (C102) Blank, T. B.; Brown, S. D. J. Chemom. 1994, 8, 391-407. SIGNAL PROCESSING (D1) Sande, G. Proc. SPIE-Int. Soc. Opt. Eng. 1993, 2029, 11527. (D2) Grunwald, E.; Herzog, J.; Steel, C. J. Chem. Educ. 1995, 72, 210-4. (D3) Graff, D. J. Chem. Educ. 1995, 72, 304-9. (D4) Ha¨bel, S.; Ga¨umann, T. Int. J. Mass Spectrom. Ion Processes 1995, 144, 139-66. (D5) Daskalova, N. Anal. Lab. 1993, 2, 79-87. (D6) Fuson, M. M. J. Chem. Educ. 1994, 71, 126-9. (D7) Levy, G. C.; Borer, P. N.; Jeong, G.-W. In Comput.-Enhance Anal. Spectrosc.; Wilkins, C. L., Ed.; Plenum: New York, 1993; Vol. 4, pp 73-96. (D8) Borer, P. N.; Levy, G. C. Methods Enzymol. 1994, 239, 25788. (D9) Nakamura, M.; Sugimoto, I.; Kuwana, H.; Lemos, R. Sens. Actuators, B 1994, 20, 231-7. (D10) Davide, F. A. M.; Di Natale, C.; D’Amico, A.; Hierlemann, A.; Mitrovics, J.; Schweizer, M.; Weimar, U.; Goepel, W.; Marco, S.; et al. Sens. Actuators, B 1995, B27, 275-85. (D11) Zhang, M.; Phillips, J. B.; J. Chromatogr., A 1995, 689, 27584. (D12) Pietrogrande, M. C.; Pasti, L.; Dondi, F.; Rodriguez, M. H. B.; Diaz, M. A. C. J. High Resolut. Chromatogr. 1994, 17, 83950. (D13) Karanassios, V.; Bateman, K. P.; Spiers, G. A. Spectrochim. Acta, Part B 1994, 49B, 989-1005. (D14) Eng, J. K.; McCormack, A. L.; Yates, J. R., III J. Am. Soc. Mass Spectrom. 1994, 5, 976-89. (D15) Doerffel, K.; Kreher, U. Fresenius’ J. Anal. Chem. 1993, 346, 516-9. (D16) Doerffel, K. Fresenius’ J. Anal. Chem. 1994, 349, 578-81. (D17) Rzhevskii, A. M.; Mardilovich, P. P. Appl. Spectrosc. 1994, 48, 13-20. (D18) Barak, P. Anal. Chem. 1995, 67, 2758-62. (D19) Kim, H. S.; Marshall, A. G. J. Mass Spectrom. 1995, 30, 123744. (D20) Alciaturi, C. E.; Escobar, M. E. Rev. Tec. Fac. Ing., Univ. Zulia 1994, 17, 1-8; Chem. Abstr. 1995, 121, 208771. (D21) Windig, W. Chemom. Intell. Lab. Syst. 1994, 23, 71-86. (D22) Riris, H.; Carlisle, C. B.; Warren, R. E.; Cooper, D. E. Opt. Lett. 1994, 19, 144-6. (D23) Riris, H.; Carlisle, C. B.; Warren, R. E.; Carr, L.; Cooper, D. E.; Martinelli, R. U.; Menna, R. J. Proc. SPIE-Int. Soc. Opt. Eng. 1995, 2366, 90-7. (D24) Mann, C. K.; Vickers, T. J. Appl. Spectrosc. 1993, 47, 19438. (D25) Szostek, B.; Trojanowicz, M. Chemom. Intell. Lab. Syst. 1994, 22, 221-8. (D26) Lee, O.; Wade, A. P.; Dumont, G. A. Anal. Chem. 1994, 66, 4507-13. (D27) Flores-Llamas, H.; Yee-Madeira, H. Nucl. Instrum. Methods Phys. Res., Sect. B 1995, 100, 176-8. (D28) Simons, J.; Bos, M.; van der Linden, W. E. Analyst 1995, 120, 1009-12. (D29) Chow, C. W. K.; Davey, D. E.; Mulcahy, D. E.; Yeow, T. C. W. Anal. Chim. Acta 1995, 307, 15-26.
(D30) Economou, A.; Fielden, P. R.; Gaydecki, P. A.; Packham, A. J. Analyst 1994, 119, 847-53. (D31) Economou, A.; Fielden, P. R. Anal. Chim. Acta 1995, 305, 165-75. (D32) Siuda, R. Appl. Surf. Sci. 1994, 81, 27-35. (D33) Ber, B. J.; Raev, A. N.; Zushinskiy, D. A. J. Electron Spectrosc. Relat. Phenom. 1994, 68, 659-67. (D34) Maxwell, J. A.; Teesdale, W. J.; Campbell, J. L. Nucl. Instrum. Methods Phys. Res., Sect. B 1995, 95, 407-21. (D35) Regehr, M. F.; Paliwal, S. K.; Regnier, F. E. J. Chromatogr. 1994, 659, 247-53. (D36) Camus, P. P.; Larson, D. J. Appl. Surf. Sci. 1994, 76-77, 41623. (D37) do Lago, C. L.; Juliano, V. F.; Kascheres, C. Anal. Chim. Acta 1995, 310, 281-8. (D38) Riris, H.; Carlisle, C. B.; Warren, R. E. Appl. Opt. 1994, 33, 5506-8. (D39) van Veen, E. H.; Bosch, S.; de Loos-Vollebregt, M. T. C. Spectrochim. Acta, Part, B 1994, 49B, 829-46. (D40) Deng, F. W.; Cushman, J. H.; Delleur, J. W. Water Resour. Res. 1993, 29, 4011-8. (D41) Cataldi, T. R. I.; Rotunno, T. Anal. Methods Instrum. 1995, 2, 27-34. (D42) Agbodjan, A. A.; Rutan, S. C. Chemom. Intell. Lab. Syst. 1994, 24, 137-44. (D43) Ferry, A.; Jacobsson, P. Appl. Spectrosc. 1995, 49, 273-8. (D44) Mosier-Boss, P. A.; Lieberman, S. H.; Newberry, R. Appl. Spectrosc. 1995, 49, 630-8. (D45) Chapados, C.; Trudel, M.; Miletic, J. Chemom. Intell. Lab. Syst. 1994, 22, 209-19. (D46) Mantsch, H. H.; Moffatt, D. J. NATO ASI Ser., Ser. C., 1993, 406, 113-24; Chem. Abstr. 1995, 121, 311383. (D47) Stockham, T. G.; Ives, J. T. U.S. Patent US 5273632, December 28, 1993. (D48) Bonnot, G.; Febvay, G. Anal. Biochem. 1995, 224, 354-63. (D49) Felinger, A. Anal. Chem. 1994, 66, 3066-72. (D50) Maudsley, A. A. J. Magn. Reson., Ser. B 1995, 106, 47-57. (D51) McCabe, M. A.; Wassall, S. R. J. Magn. Reson., Ser. B 1995, 106, 80-2. (D52) Zhang, Q.; Chen, J.; Gozansky, E. K.; Zhu, F.; Jackson, P. L.; Gorenstein, D. G. J. Magn. Reson., Ser. B 1995, 106, 164-9. (D53) Zelic, M.; Pizeta, I.; Brancia, M. Anal. Chim. Acta 1993, 281, 63-70. (D54) Labowsky, M. J. U.S. Patent US 5440119, August 8, 1995. (D55) Tsai, J. Y.; Canovai, C.; Osburn,.C. M.; Wang, Q. F.; Rose, J.; Cowen, A.; Denker, M. S. J. Vac. Sci. Technol., B 1994, 12, 219-29. (D56) Cameron, B. R.; Harland, P. W. Rev. Sci. Instrum. 1994, 65, 108-15. (D57) Meray, L. Nucl. Instrum. Methods Phys. Res., Sect. A 1994, 353, 272-5. (D58) Georgiev, A.; Gast, W. IEEE Trans. Nucl. Sci. 1993, 40, 770-9 (4, Part 1). (D59) Battistoni, C.; Mattogno, G.; Righini, G. Surf. Interface Anal. 1994, 22, 98-102. (D60) Busse, H.; Wandelt, K.; Castro, G. R. J. Electron Spectrosc. Relat. Phenom. 1995, 72, 311-6. (D61) Konkol, A.; Wilshaw, P. R.; Booker, G. R. Ultramicroscopy 1994, 55, 183-95. (D62) Gaiduk, I. M. J. Trace Microprobe Tech. 1993, 11, 271-84. (D63) Van Cappellen, E. Inst. Phys. Conf. Ser. 1993, 130, 375-8. (D64) Liew, S. C.; Loh, K. K.; Tang, S. M. Nucl. Instrum. Methods Phys. Res., Sect. B 1994, 85, 621-6. (D65) Wang, S.; Pelczer, I.; Borer, P. N.; Levy, G. C. J. Magn. Reson., Ser. A 1994, 108, 171-6. (D66) Matlengiewicz, M.; Henzel, N.; Czachowska, D.; Schmit-Quiles, F.; Nicole, D.; Lauer, J.-C. Fuel 1994, 73, 851-65. (D67) Frederick, B. G.; Nyberg, G. L.; Richardson, N. V. J. Electron Spectrosc. Relat. Phenom. 1993, 64-65, 825-34. (D68) Morawski, R. Z.; Szczecinski, L.; Barwicz, A. J. Chemom. 1995, 9, 3-20. (D69) Dobrowski, J. C.; Jamroz, M. H. J. Math. Chem. 1995, 17, 111-21. (D70) Ro, C.-U. J. Korean Chem. Soc. 1995, 39, 617-26; Chem. Abstr. 1995, 123, 211954. (D71) Vekemans, B.; Janssens, K.; Vincze, L.; Adams, F.; Van Epsen, P. Spectrochim. Acta, Part B 1995, 50B, 149-69. (D72) Bridges, F.; Booth, C. H.; Li, G. G. Physica B 1995, 208, 209, 121-4. (D73) Hill, W. Appl. Spectrosc. 1993, 47, 2171-2. (D74) Brown, D. E. J. Magn. Reson., Ser. A 1995, 114, 268-70. (D75) Levy, G. C.; Jeong, G.-Y.; Yu, J.; Wang, K. J. Magn. Reson., Ser. A 1994, 111, 179-85. (D76) Chylla, R. A.; Markley, J. L. J. Magn. Reson., Ser. B 1993, 102, 148-54. (D77) Kaljurand, M.; Urbas, E.; Haldna, U. J. Chromatogr., A 1994, 660, 147-52. (D78) Sun, X.-Y.; Singh, H.; Millier, B.; Warren, C. H.; Aue, W. A. J. Chromatogr., A 1994, 687, 259-81. (D79) Johnston, T. E. Anal. Chem. 1995, 67, 2835-41. (D80) Matthee, T.; Vissier, K. Spectrochim. Acta, Part B 1995, 50B, 823-35. (D81) Coles, B. J.; Gleeson, S. A.; Wilkinson, J. J.; Ramsey, M. H. Analyst 1995, 120, 1421-5. (D82) Salit, M. L.; Collins, J. B.; Yates, D. A. Appl. Spectrosc. 1994, 48, 915-25.
Analytical Chemistry, Vol. 68, No. 12, June 15, 1996
(D83) van Veen, E. H.; Bosch, S.; de Loos-Vollebregt, M. T. C. Spectrochim. Acta, Part B 1994, 49B, 1347-61. (D84) Roehl, R.; Gomez, J.; Woodhouse, L. R. J. Anal. At. Spectrom. 1995, 10, 15-23. (D85) Iwata, T.; Koshoubu, J. Appl. Spectrosc. 1994, 48, 1453-56. (D86) Kroutil, R. T.; Combs, R. J.; Knapp, R. B.; Small, G. W. Proc. SPIE-Int. Soc. Opt. Eng. 1993, 2089, 24-31. (D87) Kroutil, R. T.; Combs, R. J.; Knapp, R. B.; Small, G. W. Appl. Spectrosc. 1994, 48, 724-32. (D88) Mattu, M. J.; Small, G. W. Anal. Chem. 1995, 67, 2269-78. (D89) Shaffer, R. E.; Small, G. W.; Combs, R. J.; Knapp, R. B.; Kroutil, R. T. Chemom. Intell. Lab. Syst. 1995, 29, 89-108. (D90) Bangalore, A. S.; Small, G. W.; Combs, R. J.; Knapp, R. B.; Kroutil, R. T. Anal. Chim. Acta 1994, 297, 387-403. (D91) Wabuyele, B. W.; Harrington, P. de B. Anal. Chem. 1994, 66, 2047-51. (D92) Wabuyele, B. W.; Harrington, P. de B. Chemom. Intell. Lab. Syst. 1995, 29, 51-61. (D93) Wang, Z.; Dean, T.; Kowalski, B. R. Anal. Chem. 1995, 67, 2379-85. (D94) Rockwood, A. L. Rapid Commun. Mass Spectrom. 1995, 9, 103-5. (D95) Rockwood, A. L.; Van Orden, S. L.; Smith, R. D. Anal. Chem. 1995, 67, 2699-704. (D96) Barouch, E.; Coifman, R.; Ma, J.; Peckerar, M. C.; Rokhlin, V. Proc. SPIE-Int. Soc. Opt. Eng. 1993, 1924, 150-70. (D97) Madan, V. K.; Abani, M. C.; Bairi, B. R. Nucl. Instrum. Methods Phys. Res., Sect. A 1994, 343, 616-22. (D98) Doyle, M.; Chapman, B. L. W.; Balschi, J. A.; Pohost, G. M. J. Magn. Reson., Ser. B 1994, 103, 128-33. (D99) van den Boogaart, A.; Ala-Korpela, M.; Jokisaari, J.; Griffiths, J. R. Magn. Reson. Med. 1994, 31, 347-58. (D100) Lee, O.; Dumont, G. A.; Tournier, P.; Wade, A. P. Anal. Chem. 1994, 66, 971-82. (D101) Balasubramanian, K. J. Chem. Inf. Comput. Sci. 1995, 35, 5819. (D102) Goelman, G.; Leigh, J. S. J. Magn. Reson., Ser. A 1993, 105, 78-81. (D103) Szczepaniak, E.; Hornak, J. P. J. Magn. Reson., Ser. A 1993, 104, 315-20. (D104) Diaz, C.; Urquidi-Macdonald, M.; Macdonald, D. D.; Ramamurthy, A. C.; van Ooij, W. J.; Sabata, A.; Stroem, M.; Stroem, G. Corros. Control Low-Cost Reliab., Proc.-Int. Corros. Congr., 12th, 1993, 5A, 3508-18. (D105) Yamamoto, K.; Ishida, H. Spectrochim. Acta, Part A 1994, 50A, 2079-90. (D106) Bos, M.; Vrielink, J. A. M. Chemom. Intell. Lab. Syst. 1994, 23, 115-22. (D107) Stark, P. B.; Herron, M. M.; Matteson, A. Appl. Spectrosc. 1993, 47, 1820-9. (D108) Shew, S. L. U.S. Patent US 5436447, July 25, 1995. (D109) von der Linden, W. Appl. Phys. A: Mater. Sci. Process. 1995, A60, 155-65. (D110) McGown, L. B.; Hemmingsen, S. L.; Shaver, J. M.; Geng, L. Appl. Spectrosc. 1995, 49, 60-6. (D111) VanderNoot, T. J. J. Electroanal. Chem. 1995, 386, 57-63. (D112) Ritter, M.; Hutter, H.; Grasserbauer, M. Fresenius’ J. Anal. Chem. 1994, 349, 186-90. (D113) Kuzuo, R.; Tanaka, M. J. Electron Microsc. 1993, 42, 240-3. (D114) Wang, D.; Wang, H. J. Wuhan Univ. Technol., Mater. Sci. Ed. 1994, 9, 19-25; Chem. Abstr. 1995, 123, 69266. (D115) Livesey, A. K.; Smith, G. C. J. Electron Spectrosc. Relat. Phenom. 1994, 67, 439-61. (D116) Werner, W. S. M.; Smith, G. C.; Livesey, A. K. Surf. Interface Anal. 1994, 21, 38-43. (D117) Greek, L. S.; Schulze, H. G.; Blades, M. W.; Bree, A. V.; Gorzalka, B. B.; Turner, R. F. B. Appl. Spectrosc. 1995, 49, 425-31. (D118) Leclerc, G.; Pireaux, J. J. Appl. Spectrosc. 1994, 48, 973-6. (D119) Baldwin, C. A.; Alexander, P.; Gladden, L. F. J. Magn. Reson., Ser. A 1995, 112, 169-75. (D120) Costello, I.; Wells, P.; Davis, J.; Benci, N.; Skernett, D.; Davies, R. Proc. SPIE-Int. Soc. Opt. Eng. 1994, 2092, 460-9. (D121) Boehmig, S. D.; Reichl, B. M.; Eisl, M. M.; Stoeri, H. Surf. Interface Anal. 1994, 22, 304-7. (D122) Oshida, K.; Kogiso, K.; Matsubayashi, K.; Takeuchi, K.; Kobayashi, S.; Endo, M.; Dresselhaus, M. S.; Dresselhaus, G. J. Mater. Res. 1995, 10, 2507-17. (D123) Pedersen, F.; Bengtsson, E.; Nordin, B. J. Chemom. 1994, 9, 389-409. (D124) Lindgren, F.; Geladi, P.; Wold, S. J. Chemom. 1994, 8, 37789. (D125) Parnell, J. R.; Yager, P. Proc. SPIE-Int. Soc. Opt. Eng. 1995, 2388, 481-92. (D126) Dhanoa, M. S.; Lister, S. J.; Barnes, R. J. Appl. Spectrosc. 1995, 49, 765-72. (D127) Helland, I. S.; Næs, T.; Isaksson, T. Chemom. Intell. Lab. Syst. 1995, 29, 233-41. (D128) Penninckx, W.; Smeyers-Verbeke, J.; Hartmann, C.; Bourguignon, B.; Massart, D. L. Chemom. Intell. Lab. Syst. 1994, 23, 137-48. (D129) Spiegelman, C. H.; Wang, C. Y. Chemom. Intell. Lab. Syst. 1994, 23, 293-99. (D130) Marco, S.; Samitier, J.; Morante, J. R. Meas. Sci. Technol. 1995, 6, 135-42. 54R
Analytical Chemistry, Vol. 68, No. 12, June 15, 1996
(D131) Vinogradov, S. A.; Wilson, D. F. Biophys. J. 1994, 67, 204859. (D132) Cook, W. B. Proc. SPIE-Int. Soc. Opt. Eng. 1994, 2266, 51021. (D133) Smirnov, A. I.; Belford, R. L. J. Magn. Reson., Ser. A 1995, 113, 65-73. (D134) Flanigan, D. F. Appl. Opt. 1995, 34, 2636-9. (D135) Gorshkov, M. V.; Kouzes, R. T. Anal. Chem. 1995, 67, 341220. (D136) Hivert, B.; Hoummady, M.; Henrioud, J. M.; Hauden, D. Sens. Actuators, B 1994, 19, 645-8. (D137) Endres, H.-E.; Goettler, W.; Jander, H. D.; Drost, S.; Sberveglieri, G.; Faglia, G.; Perego, C. Sens. Actuators, B 1995, B25, 785-9. (D138) Endres, H.-E.; Goettler, W.; Jander, H. D.; Drost, S.; Sandmaier, H.; Sberveglieri, G.; Faglia, G.; Perego, C. Sens. Actuators, B 1995, B27, 267-70. (D139) Hayashi, Y.; Matsuda, R. Anal. Chem. 1994, 66, 2874-81. (D140) Alsberg, B. K.; Nodland, E.; Kvalheim, O. M. J. Chemom. 1994, 8, 127-45. (D141) Fu, F.-N.; DeOliveira, D. B.; Trumble, W. R.; Sarkar, H. K.; Singh, B. R. Appl. Spectrosc. 1994, 48, 1432-41. (D142) Smeller, L.; Goossens, K.; Heremans, K. Vib. Spectrosc. 1995, 8, 199-203. (D143) Fuentes, E. J.; Pachter, R.; Tonis, P. A. In Vivo 1994, 8, 199205. (D144) Lanzavecchia, S.; Tosoni, L.; Bellon, P. L. Comput. Appl. Biosci. 1995, 11, 373-8. RESOLUTION (E1) Brereton, R. G. Analyst 1995, 120, 2313-36. (E2) Quencer, B. M.; Crouch, S. R. Crit. Rev. Anal. Chem. 1993, 24, 243-62. (E3) Pitea, D.; Bonati, L.; Collina, E.; Cosentino, U.; Lasagni, M.; Moro, G. Fresenius’ J. Anal. Chem. 1994, 348, 111-20. (E4) Ehrlich, R.; Wenning R. J.; Johnson, G. W.; Su, S. H.; Paustenbach, D. J. Arch. Environ. Contam. Toxicol. 1994, 27, 486-500. (E5) Jiang, W.; Zhu, P. Sci. China, Ser. A 1993, 36, 1329-39. (E6) Wang, H.; Wang, D.; Wang, Y.; Chen, S.; Zhang, F. Analyst 1995, 120, 1603-8. (E7) Bakken, G. A.; Kalivas, J. H. Anal. Chim. Acta 1995, 300, 17381. (E8) Cuesta Sanchez, F.; Toft, J.; Kvalheim, O. M.; Massart, D. L. Anal. Chim. Acta 1995, 314, 131-9. (E9) Liang, Y.; Kvalheim, O. M. Anal. Chim. Acta 1994, 292, 5-15. (E10) Gurden, S. P.; Brereton, R. G.; Groves, J. A. Anal. Proc. 1994, 31, 205-7. (E11) Gurden, S. P.; Brereton, R. G.; Groves, J. A. Analyst 1995, 120, 1107-14. (E12) Gurden, S. P.; Brereton, R. G.; Groves, J. A. Chemom. Intell. Lab. Syst. 1994, 23, 123-35. (E13) Moekenberg, S. Fresenius’ J. Anal. Chem. 1994, 348, 496500. (E14) Windig, W. Chemom. Intell. Lab. Syst. 1994, 23, 71-86. (E15) Tauler, R. Chemom. Intell. Lab. Syst. 1995, 30, 133-46. (E16) Liang, Y.; Manne, R.; Kvalheim, O. M. Chemom. Intell. Lab. Syst. 1994, 22, 229-40. (E17) Toft, J.; Kvalheim, O. M.; Libnau, F. O.; Nodland, E. Vib. Spectrosc. 1994, 7, 125-37. (E18) Phillips, G. R.; Georghiou, S. Biophys. J. 1993, 65, 918-26. (E19) Gui, M.; Rutan, S. C.; Agbodjan, A. Anal. Chem. 1995, 67, 3293-9. (E20) Ferreira, M. M. C.; Brandes, M. L.; Ferreira, I. M. C.; Booksh, K. S.; Dolowy, W. C.; Gouterman, M.; Kowalski, B. R. Appl. Spectrosc. 1995, 49, 1317-25. (E21) Elbergali, A. K.; Brereton, R. G.; Rahmani, A. Analyst 1995, 120, 2207-16. (E22) Neal, S. L. J. Chemom. 1994, 8, 245-61. (E23) Tauler, R.; Smilde, A. K.; Kowalski, B. R. J. Chemom. 1995, 9, 31-58. (E24) Tauler, R.; Smilde, A. K.; Henshaw, J. M.; Burgess, L. W.; Kowalski, B. R. Anal. Chem. 1994, 66, 3337-44. (E25) Mitchell, B. C.; Burdick, D. S. J. Chemom. 1994, 8, 155-68. (E26) Esteves da Silva, J. C. G.; Machado, A. A. S. C. Talanta 1994, 41, 2095-104. (E27) Cuesta Sanchez, F.; Massart, D. L. Anal. Chim. Acta 1994, 298, 331-9. (E28) Norgaard, L.; Ridder, C. Chemom. Intell. Lab. Syst. 1994, 23, 107-14. (E29) Grung, B.; Kvalheim, O. M. Anal. Chim. Acta 1995, 304, 5766. (E30) Faber, N. M.; Buydens, L. M. C.; Kateman, G. J. Chemom. 1994, 8, 147-54. (E31) Faber, N. M.; Buydens, L. M. C.; Kateman, G. J. Chemom. 1994, 8, 181-204. (E32) Faber, N. M.; Buydens, L. M. C.; Kateman, G. J. Chemom. 1994, 8, 273-86. (E33) Li, S.; Gemperline, P. J.; Briley, K.; Kazmierczak, S. J. Chromatogr. 1994, 655, 213-24. (E34) Elbergali, A. K.; Brereton, R. G. Chemom. Intell. Lab. Syst. 1995, 27, 55-71. (E35) Elbergali, A. K.; Brereton, R. G. Chemom. Intell. Lab. Syst. 1994, 23, 97-106.
(E36) Elbergali, A. K.; Brereton, R. G. J. Chemom. 1994, 8, 42337. (E37) Brereton, R. G.; Gurden, S. P.; Groves, J. A. Chemom. Intell. Lab. Syst. 1995, 27, 73-87. (E38) Koel, M.; Kaljurand, M.; Lochmueller, C. H.; Moebus, M. Chemom. Intell. Lab. Syst. 1995, 30, 173-78. (E39) Grung, B.; Kvalheim, O. M. Chemom. Intell. Lab. Syst. 1995, 29, 75-87. (E40) Toft, J.; Kvalheim, O. M. Chemom. Intell. Lab. Syst. 1994, 25, 61-75. (E41) Neal, S. L. In Computer Enhanced Analytical Spectroscopy; Wilkins, C. L., Ed.; Plenum: New York, 1993; Vol. 4, pp 5471. (E42) Wentzell, P. D.; Hughes, S. G.; Vanslyke, S. J. Anal. Chim. Acta 1995, 307, 459-70. (E43) Wentzell, P. D.; Hughes, S. G. Talanta 1995, 42, 1361-71. (E44) Hayashi, Y.; Helburn, R. S.; Rutan, S. C. In Computer Enhanced Analytical Spectroscopy; Wilkins, C. L., Ed.; Plenum: New York, 1993; Vol. 4, pp 257-74. (E45) Helburn, R. S.; Rutan, S. C.; Pompano, J.; Mitchem, D.; Patterson, W. T. Anal. Chem. 1994, 66, 610-18. (E46) Alava-Moreno, F.; Liu, Y. M.; Diaz-Garcia, M. E.; Sanz-Medel, A. Mikrochim. Acta 1993, 112, 47-54. (E47) Jimenez-Prieto, R.; Velasco, A.; Silva, M.; Perez-Bendito, D. Talanta 1993, 40, 1731-9. (E48) Van den Bogaert, B.; Boelens, H. F. M.; Smit, H. C. Chemom. Intell. Lab. Syst. 1994, 25, 297-311. (E49) Reh, E. Trends Anal. Chem. 1995, 14, 1-5. (E50) Faber, N. M.; Buydens, L. M. C.; Kateman, G. Anal. Chim. Acta 1994, 296, 1-20. (E51) Faber, N. M.; Buydens, L. M. C.; Kateman, G. Chemom. Intell. Lab. Syst. 1994, 25, 203-26. (E52) Crisponi, G.; Cristiani, F.; Nurchi, V. Anal. Chim. Acta 1993, 281, 197-206. (E53) Bramanti, E.; Bramanti, M.; Stiavetti, P.; Benedetti, E. J. Chemom. 1994, 8, 409-21. (E54) Fujieda, S.; Uchida, R. Anal. Sci. 1993, 9, 701-6. (E55) de Weijer, A. P.; Lucasius, C. B.; Buydens, L. M. C.; Kateman, G.; Heuvel, H. M.; Mannee, H. Anal. Chem. 1994, 66, 2331. (E56) Broten, G. S.; Wood, H. C. Meas. Sci. Technol. 1993, 4, 1096105. (E57) de Weijer, A. P.; Buydens, L. M. C.; Kateman, G.; Heuvel, H. M. Chemom. Intell. Lab. Syst. 1995, 28, 149-64. (E58) Soltanian-Zadeh, H.; Yagle, A. E. IEEE Trans. Nucl. Sci. 1993, 40, 1204-12. (E59) Cuesta-Sanchez, F.; Khots, M. S.; Massart, D. L.; De Beer, J. O. Anal. Chim. Acta 1994, 285, 181-92. (E60) Cuesta-Sanchez, F.; Khots, M. S.; Massart, D. L. Anal. Chim. Acta 1994, 290, 249-58. CALIBRATION (F1) Thomas, E. V. Anal. Chem. 1994, 66, 795A-804A. (F2) Danzer, K. Fresenius’ J. Anal. Chem. 1995, 351, 3-10. (F3) Cerda, V.; Estela, J. M.; Forteza, R.; Cladera, A.; Gomez, E.; Oms, M. T. Int. J. Environ. Anal. Chem. 1993, 52, 159-78. (F4) Yu, R.; Xie, Y. Anal. Sci. 1994, 10, 149-53; Chem. Abstr. 1995, 120, 207417. (F5) Cummins, D. J.; Andrews, C. W. J. Chemom. 1995, 9, 489508. (F6) Zhu, E.; Barnes, R. M. J. Chemom. 1995, 9, 363-72. (F7) Ra¨nnar, S.; Lindgren, F.; Geladi, P.; Wold, S. J. Chemom. 1994, 8, 111-26. (F8) Ra¨nnar, S.; Geladi, P.; Lindgren, F.; Wold, S. J. Chemom. 1995, 9, 459-70. (F9) Lindgren, F.; Geladi, P.; Wold, S. J. Chemom. 1994, 8, 37790. (F10) Lindgren, F.; Geladi, P.; Rannar, S.; Wold, S. J. Chemom. 1994, 8, 349-64. (F11) Lindgren, F.; Geladi, P.; Berglund, A.; Sjo¨stro ¨m, M.; Wold, S. J. Chemom. 1995, 9, 331-42. (F12) Thomas, E. V. J. Chemom. 1995, 9, 471-82. (F13) Sun, J. J. Chemom. 1995, 9, 21-30. (F14) Ho ¨skuldsson, A. J. Chemom. 1995, 9, 91-124. (F15) Langsrud, Ø.; Næs, T. J. Chemom. 1995, 9, 483-88. (F16) de Jong, S.; ter Braak, C. J. F J. Chemom. 1994, 8, 169-74. (F17) de Jong, S. J. Chemom. 1995, 9, 323-6. (F18) de Jong, S.; Farebrother, R. W. Chemom. Intell. Lab. Syst. 1994, 25, 179-81. (F19) Wang, Z.; Isaksson, T.; Kowalski, B. R. Anal. Chem. 1994, 66, 249-60. (F20) Ge, Z.; Cavinato, A. G.; Callis, J. B. Anal. Chem. 1994, 66, 1354-62. (F21) DiFoggio, R. Appl. Spectrosc. 1995, 49, 67-75. (F22) Nikolyukin, A. S.; Kuz’menko, N. E. Anal. Lett. 1993, 26, 2031-60. (F23) Bermond, A. P. Analusis 1994, 22, 237-40. (F24) Herrero, A.; Ortiz, M. C.; Arcos, J.; Lopez-Palacios, J.; Sarabia, L. Anal. Chim. Acta 1994, 293, 277-93. (F25) Bechmann, I. E.; Norgaard, L.; Ridder, C. Anal. Chim. Acta 1995, 304, 229-36. (F26) Selby, M. At. Spectrosc. 1994, 15, 27-35. (F27) Boonen, S.; Verrept, P.; Moens, L. J.; Dams, R. F. J. J. Anal. At. Spectrom. 1993, 8, 711-4.
(F28) Bosch-Reig, F.; Campins-Falco, P.; Verdu-Andres, J. Anal. Chim. Acta 1993, 283, 811-30. (F29) Campins-Falco, P.; Verdu-Andres, J.; Bosch-Reig, F.; MolinsLegua, C. Anal. Chim. Acta 1995, 302, 323-33. (F30) Verdu-Andres, J; Campins-Falco, P.; Bosch-Reig, F. Talanta 1994, 41, 1569-76. (F31) Campins-Falco, P.; Verdu-Andres, J.; Bosch-Reig, F. Analyst 1994, 119, 2123-7. (F32) Campins-Falco, P.; Verdu-Andres, J.; Bosch-Reig, F. Anal. Chim. Acta 1995, 315, 267-78. (F33) Calvente, J. J.; Aparicio, R. Anal. Chim. Acta 1995, 312, 28194. (F34) Xie, Y.-L.; Liang, Y.-Z.; Jiang, J.-H.; Yu, R.-Q. Anal. Chim. Acta 1995, 313, 185-96. (F35) Greip, M. I.; Wakeling, I. N.; Vankeerberghen, P.; Massart, D. L. Chemom. Intell. Lab. Syst. 1995, 29, 37-50. (F36) Brown, P. J. Chemom. Intell. Lab. Syst. 1995, 28, 139-148. (F37) Brown, P. J. J. Chemom. 1993, 7, 255-65. (F38) Analyst 1994, 119, 2363-6. (F39) Xie, Y.-L.; Liang, Y.-Z.; Chen, Z.-G.; Huang, Z.-H.; Yu, R.-Q. Chemom. Intell. Lab. Syst. 1995, 27, 21-32. (F40) Oman, S. D.; Næs, T.; Zube, A. J. Chemom. 1993, 7, 195212. (F41) Walczak, B.; Massart, D. L. Chemom. Intell. Lab. Syst. 1995, 27, 41-54. (F42) Ruyken, M. A.; Visser, J. A.; Smilde, A. K. Anal. Chem. 1995, 67, 2170-9. (F43) Walczak, B. Chemom. Intell. Lab. Syst. 1995, 29, 63-74. (F44) Walczak, B. Chemom. Intell. Lab. Syst. 1995, 28, 259-72. (F45) Kalantar, A. H.; Gelb, R. I.; Alper, J. S. Talanta 1995, 42, 597604. (F46) Danzer, K.; Wagner, M.; Fischbacher, C. Fresenius’ J. Anal. Chem. 1995, 352, 407-12. (F47) Rajiko, R. Anal. Lett. 1994, 27, 215-28. (F48) Vankeerberghen, P.; Smeyers-Verbeke, J., Leardi, R.; Karr, C. L.; Massart, D. L. Chemom. Intell. Lab. Syst. 1995, 28, 7388. (F49) Vigneau, E.; Deveaux, M. F.; Safar, M. J. Chemom. 1995, 9, 125-35. (F50) Perez-Arribas, L. V.; Leon-Gonzales, M. E.; Polo-Diaz, L. M. Analusis 1994, 22, 369-72. (F51) Torralba, R.; Bonilla, M.; Perez-Arribas, L. V.; Palacios, A.; Camara, C. Spectrochim. Acta Part B 1994, 49B, 893-9. (F52) Mueller, A.; Staats, G.; Troebs, V. Fresenius’ J. Anal. Chem. 1994, 348, 615-25. (F53) Wienke, D.; Vijn, T.; Buydens, L. Anal. Chem. 1994, 66, 8419. (F54) Perez-Arribas, L. V.; Navarro-Villosada, F.; Leon-Gonzales, M. E.; Polo-Diaz, L. M. J. Chemom. 1993, 7, 267-75. (F55) Navarro-Villoslada, F.; Perez-Arribas, L. V.; Leon-Gonzalez, M. E.; Polo-Diaz, L. M. Anal. Chim. Acta 1995, 308, 238-45. (F56) van Veen, E. H.; Bosch, S.; de Loos-Vollebregt, M. T. C. Spectrochim. Acta, Part B 1993, 48B, 1691-701. (F57) Navarro-Villoslada, F.; Perez-Arribas, L. V.; Leon-Gonzalez, M. E.; Polo-Diaz, L. M. Anal. Chim. Acta 1995, 313, 93-102. (F58) Heikka, R.; Minkkinen, P.; Taavitsainen, V.-M. Process Control Qual. 1994, 6, 47-54. (F59) Jouan-Rimbaud, D.; Walczak, B.; Massart, D. L.; Last, I. R.; Prebble, K. A. Anal. Chim. Acta 1995, 304, 285-95. (F60) Almoey, T.; Haugland, E. Appl. Spectrosc. 1994, 48, 327-32. (F61) Keller, H. R.; Roettele, J.; Bartels, H. Anal. Chem. 1994, 66, 937-43. (F62) Geladi, P. Spec. Publ.-R. Soc. Chem. 1995, 163, 225-36. (F63) Forina, M.; Drava, G.; Boggia, R.; Lanteri, S.; Conti, P. Anal. Chim. Acta 1994, 295, 109-18. (F64) Grung, B.; Kvalheim, O. M. Chemom. Intell. Lab. Syst. 1994, 22, 115-25. (F65) de Noord, O. E. Chemom. Intell. Lab. Syst. 1994, 25, 85-97. (F66) Workman, J.; Coates, J. Spectroscopy 1993, 8(9), 36-42. (F67) Bouveresse, E.; Massart, D. L.; Dardenne, P. Anal. Chem. Acta 1994, 297, 405-16. (F68) Wang, Z.; Dean, T.; Kowalski, B. R. Anal. Chem. 1995, 67, 2379-85. (F69) Cowe, I. A.; Eddison, C. G.; Hewitt, N.; Davies, A. M. C. Spec. Pub.-R. Soc. Chem. 1995, 163, 175-83. (F70) Mestek, O.; Pavlik, J.; Suchanek, M. Fresenius’ J. Anal. Chem. 1994, 350, 344-51. (F71) Forina, M.; Drava, G.; Armanino, C.; Boggia, R.; Lanteri, S.; Leardi, R.; Corti, P.; Giangiacomo, R. Chemom. Intell. Lab. Syst. 1995, 27, 189-203. (F72) Blanco, M.; Coello, J.; Iturriaga, H.; Maspoch, S.; Rovira, E. Appl. Spectrosc. 1995, 49, 593-7. (F73) Peters, D. C.; Shearer, J. C. Proc. SPIE-Int. Soc. Opt. Eng. 1993, 2069, 32. (F74) Lebouille, J. L. M.; Drost, W. C. Monogr.-Eur. Brew. Conv. 1993, 20, 14-22. (F75) Faber, N. M.; Buydens, L. M. C.; Kateman, G. J. Chemom. 1994, 8, 181-203. (F76) Booksh, K. Kowalski, B. R. J. Chemom. 1994, 8, 45-63. (F77) Poe, R. B.; Rutan, S. C. Anal. Chim Acta 1993, 283, 845-53. (F78) Brown, S. D.; Harper, A. M. In Computer-Enhanced Analytical Spectroscopy; Wilkins, C. L., Ed.; Plenum Press: New York, 1994; Vol 4, pp 135-63. (F79) Smilde, A.; Wang, Y.; Kowalski, B. R. J. Chemom. 1994, 8, 21-36.
Analytical Chemistry, Vol. 68, No. 12, June 15, 1996
(F80) Wang, Y.; Borgen, O. S.; Kowalski, B. R. J. Chemom. 1993, 7, 439-45. (F81) Xie, Y.; Wang, J.; Liang, Y.; Ge, K.; Yu, R. Anal. Chim. Acta 1993, 281, 207-18. (F82) Booksh, K.; Henshaw, J. M.; Burgess, L. W.; Kowalski, B. R. J. Chemom. 1995, 9, 263-82. (F83) Lin, Z.; Booksh, K. S.; Burgess, L. W.; Kowalski, B. R. Anal. Chem. 1994, 66, 2552-60. (F84) Booksh, K. S.; Lin, Z.; Wang, Z.; Kowalski, B. R. Anal. Chem. 1994, 66, 2561-9. (F85) Xie, Y-L.; Baeza-Baeza, J. J.; Ramis-Ramos, G. Chemom. Intell. Lab. Syst. 1995, 27, 211-20. (F86) Miller, K. L. Analusis 1995, 23(4), M26-7. (F87) Evans, D. G.; Legrand, A.; Jewell, K.; Scotter, C. N. G. J. Near Infrared Spectrosc. 1993, 1, 209-19. (F88) De Lene Mirouze, F.; Boulou, J. C.; Dupuy, N.; Meurens, M.; Huvenne, J. P.; Legrand, P. Appl. Spectrosc. 1993, 47, 118791. (F89) Heise, H. M.; Bittner, A. J. Mol Struct. 1995, 348, 127-30. (F90) Heise, H. M.; Bittner, A. J. Mol Struct. 1995, 348, 21-4. (F91) Heise, H. M.; Marbach, R.; Koschinsky, T.; Gries, F. A. Appl. Spectrosc. 1994, 48, 85-95. (F92) Bhandare, P.; Mendelson, Y.; Stohr, E.; Peura, R. A. Vib. Spectrosc. 1994, 6, 363-78. (F93) Chung, H.; Arnold, M. A.; Rhiel, M.; Murhammer, D. W. Appl. Biochem. Biotechnol. 1995, 50, 109-25. (F94) Bhandare, P.; Mendelson, Y.; Stohr, E.; Peura, R. A. Appl. Spectrosc. 1994, 48, 271-3. (F95) Simhi, R.; Bunimovich, D.; Sela, B.-A.; Katzir, A. Proc. SPIEInt. Soc. Opt. Eng. 1995, 2388, 493-500. (F96) Hall, J. W.; Pollard, A. Clin. Biochem. 1993, 26, 483-90. (F97) Hazen, K. H.; Arnold, M. A.; Small, G. W. Appl. Spectrosc. 1994, 48, 477-83. (F98) Heise, H. M.; Marbach, R. Proc SPIE-Int. Soc. Opt. Eng. 1993, 2089, 114-5. (F99) Haaland, D. M.; Jones, H. D. T. Proc. SPIE-Int. Soc. Opt. Eng. 1993, 2089, 448-9. (F100) Bittner, A.; Marbach, R.; Heise, H. M. J. Mol. Struct. 1995, 349, 341-4. (F101) Hall, J. W.; Pollard, A. J. Near Infrared Spectrosc. 1993, 1, 12732. (F102) Domjan, G.; Kaffka, K. J.; Jako, J. M. Valyi-Nagy, I. T. J. Near Infrared Spectrosc. 1994, 2, 67-78. (F103) Brimmer, P. J.; Hall, J. W. Can. J. Appl. Spectrosc. 1993, 38, 155-62. (F104) Jalsovsky, G.; Egyed, O.; Holly, S.; Hegedus, B. Appl. Spectrosc. 1995, 49, 1142-5. (F105) van de Voort, F. R.; Ismail, A. A.; Sedman, J.; Dubois, J.; Nicodemo, T. J. Am. Oil Chem. Soc. 1994, 71, 921-6. (F106) Roy, S.; Anantheswaran, R. C.; Shenk, J. S.; Westerhaus, M. O.; Beelman, R.B. J. Sci. Food Agric. 1993, 63, 355-60. (F107) Isaksson, T.; Toegersen, G.; Iversen, A.; Hildrum, K. I. J. Sci. Food Agric. 1995, 69, 95-100. (F108) Delwiche, S. R. Cereal Chem. 1995, 72, 11-6. (F109) Bak, J.; Larsen, A. Appl. Spectrosc. 1995, 49, 437-43. (F110) Hughes, T. L.; Methven, C. M.; Jones, T. G. J.; Pelham, S. E.; Fletcher, P.; Hall, C. Adv. Cem. Based Mater. 1995, 2, 91104. (F111) Janik, L. J.; Skjemstand, J. O.; Raven, M. D. Aust. J. Soil Res. 1995, 33, 621-36. (F112) Janik, J. J.; Skjemstand, J. O. Aust. J. Soil Res. 1995, 33, 63750. (F113) Franke, J. E.; Niemczyk, T. M.; Haaland, D. M. Spectrochim. Acta, Part A 1994, 50A, 1687-723. (F114) Franke, J. E.; Zhang, L.; Niemczyk, T. M.; Haaland, D. M.; Radigan, K. J. J. Vac. Sci. Technol. A 1995, 13, 1959-66. (F115) Berger, A. J.; Wang, Y.; Sammeth, D. M.; Itzkan, I.; Kneipp, K.; Feld, M. S. Appl. Spectrosc. 1995, 49, 1164-9. (F116) Gossen, P. D.; MacGregor, J. F.; Pelton, R. H. Appl. Spectrosc. 1993, 47, 1852-70. (F117) Blanco, M.; Coello, J.; Iturriaga, H.; Maspoch, S.; de la Pezuela, C.; Russo, E. Anal. Chim. Acta 1994, 298, 183-91. (F118) Andrew, K. N.; Worsfold, P. J. Analyst 1994, 119, 1541-6. (F119) Yan, H. M.; Kraus, G.; Gauglitz, G. Anal. Chim. Acta 1995, 312, 1-8. (F120) Havel, J.; Jimenez, F.; Bautista, R. D.; Leon, J. J. A. Analyst 1993, 118, 1355-60. (F121) Chang, K.; Brown, C. W.; Force, R. K. Appl. Spectrosc. 1993, 47, 1801-7. (F122) Danzer, K.; Venth, K. Fresenius’ J. Anal. Chem. 1994, 350, 339-43. (F123) Schultze, D.; Stilbs, P. J. Mag. Reson. Ser. A 1993, 105, 54-8. (F124) Hanning, A. Int. Patent Appl. WO 9509355, April 6, 1995. (F125) Norli, H. R.; Esbensen, K.; Westad, F.; Birkeland, K. I.; Hemmersbach, P. J. Steroid Biochem. Mol. Biol. 1995, 54, 838. (F126) Sato, M.; Mitsu, T. Anal. Sci. 1994, 10, 485-90; Chem. Abstr. 1995, 121, 148072. (F127) Jagner, D.; Renman, L.; Stefansdottir, S. H. Electroanalysis 1994, 6, 201-8. (F128) Hierlemann, A.; Weimar, U.; Kraus, G.; Gauglitz, G.; Goepel, W. Sens. Mater. 1995, 7, 179-89. (F129) Pribic, R. Anal. Biochem. 1994, 223, 26-34. (F130) Lew, R.; Balke, S. T. Appl. Spectrosc. 1993, 47, 1747-50. (F131) Chilkoti, A.; Schmierer, A. E.; Perez-Luna, V. H.; Ratner, B. D. Anal. Chem. 1995, 67, 2883-91. 56R
Analytical Chemistry, Vol. 68, No. 12, June 15, 1996
(F132) Niemczyk, T. M.; Franke, J. E.; Wangmaneerat, B.; Chen, C. S.; Haaland, D. M. In Computer Enhanced Analytical Spectroscopy; Wilkins, C. L., Ed.; Plenum: New York, 1993; pp 16585. (F133) Niemczyk, T. M.; Wangmaneerat, B.; Haaland, D. M. J. Vac. Sci. Technol. A 1994, 12, 835-8. (F134) Everall, N. J.; Chalmers, J. M.; Ferwerda, R.; van der Maas, J. H.; Hendra, P. J. J. Raman Spectrosc. 1994, 25, 43-51. (F135) Fredstrom, S. B.; Jung, H. J. G.; Halgerson, J. L.; Eyden, C. A.; Slavin, J.L. J. Agric. Food Chem. 1994, 42, 735-8. (F136) Garcia-Jares, C.; Medina, B. J. Sci. Food Agric. 1993, 63, 34954. (F137) Luning, P. A.; van der Vuurst de Vries, R.; Yuksel, D.; Ebbenhorst-Seller, T.; Wichers, H. J.; Roozen, J. P. J. Agric. Food Chem. 1994, 42, 2855-61. (F138) Letzelter, N. S.; Wilson, R. H.; Jones, A. D.; Sinnaeve, G. J. Sci. Food Agric. 1995, 67, 239-45. (F139) Agblevor, F. A.; Evans, R. J.; Johnson, K. D. J. Anal. Appl. Pyrolysis 1994, 30, 125-44. (F140) Lennholm, H.; Iversen, T. Holzforchung 1995, 49, 119-26. (F141) Lennholm, H.; Wallbaecks, L.; Iversen, T. Nord. Pulp Pap. Res. J. 1995, 10, 46-50. (F142) Nilsson, M.; Korsman, T.; Nordgren, A.; Palmborg, C.; Renberg, I.; O ¨ hman, J. In Near Infra-Red Spectroscopy; Hildrum, K. I., Ed.; Horwood: Chichester, UK, 1992; pp 229-34. (F143) Meuret, M.; Dardenne, P.; Biston, R.; Poty, O. J. Near Infrared Spectrosc. 1993, 1, 45-54. (F144) De Boever, J. L.; Cottyn, B. G.; Vanacker, J. M.; Boucque, Ch. V. Anim. Feed Sci. Technol. 1995, 51, 243-53. (F145) De Boever, J. L.; Eeckhout, W.; Boucque, Ch.V. Neth. J. Agric. Sci. 1994, 42, 357-69. (F146) Aber, J. D.; Bolster, K. L.; Newman, S. D.; Soulia, M.; Martin, M. E. J. Near Infrared Spectrosc. 1994, 2, 15-23. (F147) Michell, A. J. Appita J. 1994, 47, 29-37; Chem Abstr. 1995, 122, 58484. (F148) de Bakker, C. J.; Fredericks, P. M. Appl. Spectrosc. 1995, 49, 1766-71. (F149) Cooper, J. B.; Wise, K. L.; Groves, J.; Welch, W. T. Anal. Chem. 1995, 67, 4096-100. (F150) Selzer, E.; Williams, R. Proc.-Beltwide Cotton Conf. 1993, 2, 1154; Chem. Abstr. 1995, 120, 325604. (F151) Jensen, B.; Wolkoff, P.; Wilkins, C. K. Indoor Air 1995, 5, 449. (F152) Wang, Z.; Hwang, J.-N.; Kowalski, B. R. Anal. Chem. 1995, 67, 1497-1504. (F153) Næs, T.; Kvaal, K.; Isaksson, T.; Miller, C. J. Near Infrared Spectrosc. 1993, 1, 1-11. (F154) Gemperline, P. J. Spec. Pub.-R. Soc. Chem. 1995, 163, 21824. (F155) Tetko, I. V.; Livingstone, D. L.; Luik, A. I. J. Comp. Inf. Comp. Sci. 1995, 35, 826-33. (F156) Zupan, J.; Novic, M.; Gasteiger, J. Chemom. Intell. Lab. Syst. 1995, 27, 175-87. (F157) Walczak, B.; Wegscheider, W. Anal. Chim. Acta 1993, 283, 508-17. (F158) Walczak. B.; Wegscheider, W. Chemom. Intell. Lab. Syst. 1994, 22, 199-207. (F159) Ni, Y.; Peng, Z. Anal. Chim. Acta 1995, 304, 217-22. (F160) Simutis, R.; Havlik, I.; Luebbert, A. IFAC Symp. Ser. 1992, 10, 95-100. (F161) Liu, D.; Tan, H.; Bao, L.; Nie, L.; Yao, S. Analyst 1995, 120, 2573-8. (F162) Robison-Cox, J. F. J. Immunol. Methods 1995, 186, 79-88. (F163) Borgaard, C. Spec. Pub.-R. Soc. Chem. 1995, 163, 209-17. (F164) Borgaard, C. Spectrosc. Eur. 1994, 6(3), 21-7. (F165) Carlin, M.; Kavli, T.; Lillekjendlie, B. Chemom. Intell. Lab. Syst. 1994, 23, 163-77. (F166) Jacobsson, S. P.; Hagman, A. Anal. Chim. Acta 1993, 284, 137-47. (F167) Wang, C. D.; Walter, W. T.; Kagann, R. H. Proc. SPIE-Int. Soc. Opt. Eng. 1995, 2366, 251-62. (F168) Jagemann, K.-U.; Fischbacher, C.; Danzer, K.; Mueller, U. A.; Mertes, B. Z. Phys. Chem. (Munich) 1995, 191, 179-90. (F169) Bhandare, P.; Mendelson, Y.; Peura, R. A.; Janatsch, G.; KruseJarres, J. D.; Marbach, R.; Heise, H. M. Appl. Spectrosc. 1993, 47, 1214-21. (F170) Fuller, M. E.; Cherne, F. J. Int. Patent Appl. WO 9504496, February 16, 1995. (F171) Veng-Pedersen, P.; Modi, N. B. J. Pharm. Sci. 1993, 82, 91826. (F172) Goodacre, R.; Karim, A.; Kaderbhai, M. A.; Kell, D. B. J. Biotechnol. 1994, 34, 185-93. (F173) Goodacre, R.; Neal, M. J.; Kell, D. B. Anal. Chem. 1994, 66, 1070-85. (F174) Goodacre, R.; Trew, S.; Wrigley-Jones, C.; Neal, M. J.; Maddock, J.; Ottley, T. W.; Porter, N.; Kell, D. B. Biotechnol. Bioeng. 1994, 44, 1205-16. (F175) Goodacre, R.; Edmonds, A. N.; Kell, D. B. J. Anal. Appl. Pyrolysis 1993, 26, 93-114. (F176) Goodacre, R.; Trew, S.; Wrigley-Jones, C.; Saunders, G.; Neal, M. J.; Porter, N.; Kell, D. B. Anal. Chim. Acta 1995, 313, 2543. (F177) Schweizer-Berberich, M.; Goeppert, J.; Hierlemann, A.; Mitrovics, J.; Weimar, U.; Rosenstiel, W.; Goepel, W. Sens. Actuators B 1995, B27, 232-6.
(F178) Hierlemann, A.; Weimar, U.; Krauss, G.; Schwizer-Berberich, M.; Goepel, W. Sens. Actuators B 1995, B26, 126-34. (F179) Moore, S. W.; Hines, E. L.; Goepel, W.; Weimar, U. Sens. Actuators B 1993, 16, 344-8. (F180) Boger, Z.; Karpas, Z. J. Chem. Inf. Comput. Sci. 1994, 34, 57680. (F181) Catasus, M.; Branagh, W.; Salin, E. D. Appl. Spectrosc. 1995, 49, 798-807. (F182) Jasper, W. J.; Kovacs, E. T.; Berstrasser, G. A. Text. Res. J. 1993, 63, 545-51. (F183) Marjoniemi, M. J. Am. Leather Chem. Assoc. 1994, 82, 3948. (F184) Marjoniemi, M. Appl. Spectrosc. 1994, 48, 21-6. (F185) van Leeuwen, J. A.; Jonker, R. J.; Gill, R. Chemom. Intell. Lab. Syst. 1994, 25, 325-39. (F186) Tolchard, J. M.; Boyd, A. Int. Patent Appl. WO 9417391, August 4, 1994. (F187) Albiol, J.; Cammajo, C.; Casas, C.; Poch, M. Biotechnol. Prog. 1995, 11, 88-92. (F188) Majcen, N.; Rajer-Kanduc, K.; Novic, M.; Zupan, J. Anal. Chem. 1995, 67, 2154-61. (F189) Vallejo-Cordoba, B.; Arteaga, G. E.; Nakai, S. J. Food Sci. 1995, 60, 885-8. (F190) Volmer, M.; Wolthers, B. B.; Metting, H. J.; de Haan, T. H. Y.; Coenegracht, P. M. J.; van der Slik, W. Clin. Chem. 1994, 40, 1692-7. (F191) Fichtner, W.; Kaden, H.; Herrmann, S. German Patent DE 4341597, June 8, 1995. PARAMETER ESTIMATION (G1) Cumpson, P. J. J. Electron Spectrosc. Relat. Phenom. 1995, 73, 25-52. (G2) Love, M. D.; Pardue, H. L. Anal. Chim. Acta 1994, 299, 195208. (G3) Hughes, S. G.; Taylor, E. L.; Wentzell, P. D.; McCurdy, R. F.; Boss, R. K. Anal. Chem. 1994, 66, 830-5. (G4) Carlin, M.; Kavli, T.; Lillekjendlie, B. Chemom. Int. Lab. Syst. 1994, 23, 163-77. (G5) Roy, T. J. Chemom. 1994, 8, 37-44. (G6) Olive, J.; Grimalt, J. O. J. Chromatogr. Sci. 1995, 33, 194203. (G7) Bauchspiess, K. R. Jpn. J. Appl. Phys., Part 1 1993, 32, 1313. (G8) Karolczak, M. Curr. Sep. 1995, 13, 98-104. (G9) Tan, H. K. S. J. Chem. Soc., Faraday Trans. 1994, 90, 35215. (G10) Koehl, P.; Ling, C.; Lefevre, J. F. J. Magn. Reson., Ser. A 1994, 109, 32-40. (G11) Martin, Y. L. J. Magn. Reson., Ser. A 1994, 111, 1-10. (G12) Van Huffel, S.; Chen, H.; Decanniere, C.; Van Hecke, P. J. Magn. Reson., Ser. A 1994, 110, 228-37. (G13) Yang, J. X.; Havel, T. F. J. Biomol. NMR 1994, 4, 807-26. (G14) Leclerc, G.; Pireaux, J. J. J. Electron Spectrosc. Relat. Phenom. 1995, 71, 141-64. (G15) Leclerc, G.; Pireaux, J. J. J. Electron Spectrosc. Relat. Phenom. 1995, 71, 165-78. (G16) Leclerc, G.; Pireaux, J. J. J. Electron Spectrosc. Relat. Phenom. 1995, 71, 179-90. (G17) Potvin, P. G. Anal. Chim. Acta 1994, 299, 43-57. (G18) Orti-Fernandez, M. C.; Herrero-Gutierrez, A. Chemom. Intell. Lab. Syst. 1995, 27, 231-43. (G19) Power, J. F.; Prystay, M. C. Appl. Spectrosc. 1995, 49, 72546. (G20) Altenburg, H. P. Eur. J. Clin. Chem. Clin. Biochem. 1994, 32, 873-80. (G21) Biader Ceipidor, U.; Cataldi, T. R. I.; Desimoni, E.; Salvi, A. M. J. Chemom. 1994, 8, 221-39. (G22) Benner, D. C.; Rinsland, C. P.; Devi, V. M.; Smith, M. A. H.; Atkins, D. J. Quant. Spectrosc. Radiat. Transfer 1995, 53, 70521. (G23) Bugnon, P.; Chottard, J. C.; Jestin, J. L.; Jung, B.; Laurenczy, G.; Maeder, M.; Merbach, A. E.; Zuberbuehler, A. D. Anal. Chim. Acta 1994, 298, 193-201. (G24) Koons, J. M.; Hughes, E.; Ellis, P. D. Anal. Chim. Acta 1993, 283, 1045-58. (G25) Yoon, Y. G.; Phun, S. I. Electrochim. Acta 1995, 40, 9991004. (G26) Wold, S. Chemom. Intell. Lab. Syst. 1994, 23, 149-61. (G27) Ho ¨skuldsson, A. J. Chemom. 1995, 9, 91-124. (G28) Zhou, H.; Wang, Y.; Ho, W. J. Vac. Sci. Technol., A 1995, 13, 2013-7. (G29) Gonza´lez, A. G.; Gonza´lez-Arjona, D. Anal. Chim. Acta 1995, 312, 295-306. (G30) Gonza´lez, A. G.; Herrador, M. A.; Asuero, A. G. Int. J. Pharm. 1995, 120, 215-20. (G31) Armanino, C.; Forina, M.; Bonfanti, L.; Maspero, M. Anal. Chim. Acta 1993, 284, 79-89. (G32) Perez-Luna, V. H.; Horbett, T. A.; Ratner, B. D. J. Biomed. Mater. Res. 1994, 28, 1111-26. (G33) Evilia, R. F.; Effiong, R.; Whittenburg, S. L. Spectrosc. Lett. 1993, 26, 1559-70. (G34) Xing, D.; Gibbs, S.; Derbyshire, J. A.; Fordham, E. J.; Carpenter, T. A.; Hall, L. D. J. Magn. Reson., Ser. B 1995, 106, 1-9. (G35) Sivia, D. S.; David, W. I. F. Acta Crystallogr., Sect. A 1994, 50, 703-25.
(G36) Joergensen, J. S.; Pedersen, J. B. Chemom. Intell. Lab. Syst. 1994, 22, 25-35. (G37) Chylla, R. A.; Markley, J. L. J. Biomol. NMR 1993, 3, 51533. (G38) Chylla, R. A.; Markley, J. L. J. Biomol. NMR 1995, 5, 24558. (G39) Prettyman, T. H.; Cole, R. A.; Estep, R. J.; Sheppard, G. A. Nucl. Instrum. Methods Phys. Res., Sect. A 1995, 356, 470-5. (G40) Brink, A.; Westerlund, T. Chemom. Intell. Lab. Syst. 1995, 29, 29-36. (G41) Stanley, B. J.; Topper, K.; Marshall, D. B. Anal. Chim. Acta 1994, 287, 25-34. (G42) Zupan, J.; Novic, M.; Gasteiger, J. Chemom. Intell. Lab. Syst. 1995, 27, 175-88. (G43) Gorburu, J. V. S.; Shelver, W. L.; Shelver, W. H. J. Liq. Chromatogr. 1995, 18, 1957-72. (G44) Ventura, S.; Silva, M.; Perez-Bendito, D.; Hervas, C. Anal. Chem. 1995, 67, 1521-5. (G45) Khadem, M.; Ipakchi, A.; Alexandro, F. J.; Colley, R. W. Proc. Am. Power Conf. 1994, 56, 772-7. (G46) Walczak, B.; Wegscheider, W. Anal. Chim. Acta 1993, 283, 508-17. (G47) Miyashita, Y.; Yoshida, H.; Yaegashi, O.; Kimura, T.; Nishiyama, H.; Sasaki, S. THEOCHEM 1994, 117, 241-5. (G48) Collette, T. W.; Szladow, A. J. Appl. Spectrosc. 1994, 48, 137986. (G49) Wu, X.; Bellgardt, K. H. Anal. Chim. Acta 1995, 313, 16176. (G50) Parczewski, A.; Lucasius, C. B.; Kateman, G. Fresenius’ J. Anal. Chem. 1994, 348, 626-32. (G51) Ellis, M. F.; Taylor, T. W.; Jensen, K. F. AIChE J. 1994, 40, 445-62. (G52) Gui, M.; Rutan, S. C. Anal. Chem. 1994, 66, 1513-9. (G53) Felinger, A.; Pap, T.; Incze´dy, J. Talanta 1994, 41, 1119-26. (G54) Li, Z.; Miyashita, Y.; Sasaki, S. In Comput. Aided Innovation New Mater. 2; Doyama, M., Ed.; North-Holland: Amsterdam, Netherlands, 1993; pp 935-8. (G55) Losev, A. Appl. Spectrosc. 1994, 48, 1289-90. (G56) Morton, D. W.; Young, C. L. J. Chromatogr. Sci. 1995, 33, 514-24. (G57) Parczewski, A.; Kateman, G.; Chemom. Intell. Lab. Syst. 1994, 25, 265-83. (G58) Piepmeier, E. H.; Peterson, K. Anal. Chem. 1994, 66, 223-9. (G59) Lupu, M.; Todor, D. Chemom. Intell. Lab. Syst. 1995, 29, 117. (G60) Chen, H.; Van Huffel, S.; Decanniere, C.; Van Hecke, P. J. Magn. Reson., Ser. A 1994, 109, 46-55. (G61) Phillips, B.; Moyers, R.; Lay, L. T. Proc. SPIE-Int. Soc. Opt. Eng. 1995, 2365, 374-88. (G62) Majer, M. J. Chemom. 1995, 9, 59-66. (G63) Vigneau, E.; Devaux, M. F.; Safar, M. J. Chemom. 1995, 9, 125-36. STRUCTURE/ACTIVITY RELATIONSHIPS (H1) Randic, M.; Mihalic, Z.; Nikolic, S.; Trinajstic, N. J. Chem. Inf. Comput. Sci. 1994, 34(2), 403-9. (H2) Okano, N.; Murakami, M.; Miyamoto, Y.; Koizumi, K.; Ogawa, H.; Wakabayaski, K. J. Pestic. Sci. 1993, 18(4), 361-8. (H3) Tanaka, A.; Nakamura, K.; Nakanishi, I.; Fujiwara, H. J. Med. Chem. 1994, 37(26), 4563-6. (H4) Nakamura, K.; Hayashi, K.; Ueda, I.; Fujiwara, H. Chem. Pharm. Bull. 1995, 43(3), 369-73. (H5) Sicbaldi, F.; Finizio, A. Mobility and Degradation of Xenobiotics, Proceedings of Symposium on Pesticide Chemistry, 9th; Del Re, A., Ed.; G. Biagini: Lucca, Italy, 1993; pp 321-34. (H6) Hatrik, S.; Lehotay, J.; Cizmarik, J. Collect. Czech. Chem. Commun. 1995, 60(6), 960-5. (H7) Gago, F.; Pastor, M.; Perez-Butragueno, J.; Lopez, R.; AlvarezBuilla, J.; Elguero, J. Quant. Struct.-Act. Relat. 1994, 13(2), 165-71. (H8) Biagi, G. L.; Recanatini, M.; Barbaro, A. M.; Guerra, M. C.; Sapone, A.; Borea, P. A.; Pietrogrande, M. C. New Approaches in Chromatography ’93; [Pap. Budapest Chromatogr. Conf.]; Meeting Date 1990, Kalasz, H., Ettre, L. S., Eds.; Pick, Judit. Fekete Sas Kiado: Budapest, Hungary, 1993; pp 215-34. (H9) Hocart, S. J.; Reddy, V.; Murphy, W. A.; Coy, D. H. J. Med. Chem. 1995, 38(11), 1974-89. (H10) Briens, F.; Bureau, R.; Rault, S.; Robba, M. SAR QSAR Environ. Res. 1994, 2(3), 147-57. (H11) Belvisi, L.; Bravi, G.; Scolastico, C.; Vulpetti, A.; Salimbeni, A.; Todeschini, R. J. Comput.-Aided Mol. Des. 1994, 8(2), 21120. (H12) Cruciani, G.; Watson, K. A. J. Med. Chem. 1994, 37(16), 2589601. (H13) Bjoersvik, H.-R.; Priebe, H. Acta Chem. Scand. 1995, 49 (6), 446-56. (H14) Musumarra, G.; Stella, M. Quant. Struct.-Act. Relat. 1993, 12(3), 256-60. (H15) Spraul, M.; Neidig, P.; Klauck, U.; Kessler, P.; Holmes, E.; Nicholson, J.K.; Sweatman, B. C.; Salman, S. R.; Farrant, R. D.; et al. J. Pharm. Biomed. Anal. 1994, 12(10), 1215-25. (H16) Mahfouz, N. M.; Moghazy, S. M. Alexandria J. Pharm. Sci. 1995, 9(1), 21-7.
Analytical Chemistry, Vol. 68, No. 12, June 15, 1996
(H17) Cruciani, G.; Clementi, S. Methods Princ. Med. Chem. 1995, 3 (Advanced Computer-Assisted Techniques in Drug Discovery), 61-88. (H18) Cocchi, M.; Menziani, M. C.; Fanelli, F; de Benedetti, P. G. THEOCHEM 1995, 331(1-2), 79-93. (H19) Belvisi, L.; Bonati, L.; Bravi, G.; Pitea, D.; Salimbeni, A.; Scolastico, C.; Todeschini, R.; Vulpetti, A. Trends in QSAR and Molecular Modelling 92, Proc. Eur. Symp. Struct.-Act. Relat., 9th; Wermuth, C.-G., Ed.; ESCOM: Leiden, The Netherlands, 1993; pp 312-3. (H20) Baroni, M.; Costantino, G.; Cruciani, G.; Riganelli, D.; Valigi, R.; Clementi, S. Trends in QSAR and Molecular Modelling 92, Proc. Eur. Symp. Struct.-Act. Relat., 9th; Wermuth, C.-G., Ed.; ESCOM: Leiden, The Netherlands, 1993; pp 320-3. (H21) Hasegawa, K.; Deushi, T.; Yoshida, H.; Miyashita, Y.; Sasaki, S. J. Comput.-Aided Mol. Des. 1994, 8(4), 449-56. (H22) Eriksson, L.; Berglind, R.; Sjoestroem, M. Chemom. Intell. Lab. Syst. 1994, 23(1), 235-45. (H23) Tosato, M. L.; Pino, A.; Passerini, L.; Marchini, S.; Vigano, L.; Hoglund, M. D. Sci. Total Environ. 1993, (Suppl., Part 2) 1479-90. (H24) Waterbeemd, H. van de Methods Princ. Med. Chem. 1995, 2, 1-13. (H25) Clementi, S.; Curciani, G.; Baroni, M.; Costantino, G. 3D QSAR Drug Design; Kubinyi, H., Ed.; ESCOM: Leiden, The Netherlands, 1993; pp 567-82. (H26) Seydel, J. K.; Cordes, H. P.; Albores-Velasco, M.; Visser, K.; Wiese, M. Trends in QSAR and Molecular Modelling 92, Proc. Eur. Symp. Struct.-Act. Relat. 9th; Wermuth, C.- G., Ed.; ESCOM: Leiden, The Netherlands, 1993; pp 120-3. (H27) Dunn, W. J., III; Wold, S. Methods Princ. Med. Chem. 1995, 2, 179-93. (H28) Waterbeemd, H. van de Methods Princ. Med. Chem. 1995, 2, 283-93. (H29) Franke, R.; Gruska, A. Methods Princ. Med. Chem. 1995, 2, 113-63. (H30) Miyashita, Y.; Li, Z.; Sasaki, S. Trends Anal. Chem. 1993, 12(3), 50-60. (H31) Lindberg, N. O.; Lundstedt, T. Drug Dev. Ind. Pharm. 1995, 21(9), 987-1007. (H32) Manallack, D. T.; Livingstone, D. J. Methods Princ. Med. Chem. 1995, 3, 293-318. (H33) Calleja, M. C.; Geladi, P.; Persoone, G. SAR QSAR Environ. Res. 1994, 2(3), 193-234. (H34) Kaliszan, R. J. Chromatogr. 1993, 656(1-2), 417-35. (H35) Woloszyn, T. F. Computer-assisted studies in quantitative structure retention relationships. Report, Order No. ADA251901, NTIS. Gov. Rep. Announce. Index (U.S.) 1992, 92(19). (H36) Baba, Y. J. Chromatogr., Biomed. Appl. 1993, 618(1-2), 4155. (H37) Kaliszan, R.; Nasal, A.; Turowski, M.; Radwanska, A.; Bober, L. In International Symposium on Chromatography, 35th Anniv. Res. Group Liq. Chromatogr. Jpn.; Hatano, H., Hanai, T., Eds.; World Scientific: Singapore, Singapore, 1995; pp 651-62. (H38) Kaliszan, R. Chemom. Intell. Lab. Syst. 1994, 24(2), 89-97. (H39) Forgacs, E.; Cserhati, T. J. Chromatogr., A 1995, 696(2), 26572. (H40) Cserhati, T.; Hollo, J. Biochem. Mol. Biol. Int. 1994, 32(2), 201-11. (H41) Forgacs, E.; Cserhati, T. J. Chromatogr., B: Biomed. Appl. 1994, 656(1), 233-8. (H42) Cserhati, T.; Forgacs, E. J. Chromatogr., A 1995, 699(1 + 2), 285-90. (H43) Cserhati, T,; Forgacs, E. J. Chromatogr., A 1994, 685(2), 295302. (H44) Cserhati, T. Int. J. Pharm., 1995, 124(2), 205-11. (H45) Forgacs, E. J. Pharm. Biomed. Anal., 1995, 13(4/5), 52532. (H46) Cserhati, T. Anal. Chim. Acta, 1994, 292(1-2), 17-22. (H47) Cserhati, T. Biomed. Chromatogr., 1994, 8(6), 267-72. (H48) Cserhati, T,; Forgacs, E. J. Chromatogr., A 1994, 665(1), 1725. (H49) Forgacs, E.; Cserhati, T. Quant. Struct.-Act. Relat., 1994, 13(1), 38-42. (H50) Forgacs, E.; Cserhati, T. J. Chromatogr., B: Biomed. Appl., 1995, 664(1), 227-85. (H51) Darwish, Y.; Cserhati, T.; Forgacs, E. J. Planar Chromatogr.Mod. TLC, 1993, 6(6), 458-62. (H52) Darwish, Y.; Cserhati, T.; Forgacs, E. Chemom. Intell. Lab. Syst., 1994, 24(2), 169-76. (H53) Pyka, A. J. Planar Chromatogr.-Mod. TLC, 1994, 7(1), 41-9. (H54) Bazylak, G. J. Planar Chromatogr.-Mod. TLC, 1994, 7(3), 20210. (H55) Amic, D.; Davidovic-Amic, D.; Trinajstic, N. J. Chem. Inf. Comput. Sci., 1995, 35(1), 136-9. (H56) Jalali-Heravi, M.; Garkani-Nejad, Z. J. Chromatogr., 1993, 648(2), 389-93. (H57) Corbella, R.; Rodriguez, M. A.; Sanchez, M. J.; Garcia Montelongo, F. Chromatographia, 1995, 40(9/10), 532-8. (H58) Katritzky, A. R.; Ignatchenko, E. S.; Barcock, R. A.; Lobanov, V. S.; Karelson, M. Anal. Chem., 1994, 66(11), 1799-807. (H59) Rodriguez Delgado, M. A.; Sanchez, M. J.; Gonzalez, V.; Garcia Montelongo, F. J. Chromatogr., A 1995, 697(1 + 2), 71-80. (H60) Bruchmann, A.; Zinn, P.; Haffer, C. M. Anal. Chim. Acta, 1993, 283(2), 869-80. 58R
Analytical Chemistry, Vol. 68, No. 12, June 15, 1996
(H61) Cupid, B. C.; Nicholson, J. K.; Davis, P.; Ruane, R. J.; Wilson, I. D.; Glen, R. C.; Rose, V. S.; Beddell, C. R.; Lindon, J. C. Chromatographia, 1993, 37(5-6), 241-9. (H62) Kim, H. S.; Kim, T. K.; Lee, D. W. J. Liq. Chromatogr., 1994, 17(12), 2615-23. (H63) Bemgard, A.; Colmsjoe, A.; Wrangskog, K. Anal. Chem., 1994, 66(23), 4288-94. (H64) Zou, H.; Zhang, Y.; Hong, M.; Lu, P. Liq. Chromatogr., 1993, 16(16), 3433-43. (H65) Forgacs, E.; Tibor, C. Analyst, 1995, 120(7), 1941-4. (H66) Forgacs, E.; Cserhati, T. J. Pharm. Biomed. Anal. 1992, 10(10-12), 861-5. (H67) Forgacs, E. J. Liq. Chromatogr. 1993, 16(17), 3757-72. (H68) Forgacs, E.; Cserhati, T. J. Chromatogr., A 1994, 668(2), 395402. (H69) Law, B.; Weir, S. J. Chromatogr. 1993, 657(1), 17-24. (H70) Timerbaev, A. R.; Semenova, O. P.; Tsoi, I. G.; Petrukhin, O. M. J. Chromatogr. 1993, 648(2), 307-14. (H71) Greenaway, M.; Okafo, G.; Manallack, D.; Camilleri, P. Electrophoresis 1994, 15(10), 1284-9. (H72) Timerbaev, A. R.; Semenova, O. P. J. Chromatogr., A 1995, 690(1), 141-8. (H73) Forgacs, E. J. Liq. Chromatogr. 1993, 16(12), 2483-500. (H74) Bazylak, G. J. Planar Chromatogr.-Mod. TLC 1994, 7(6), 42834. (H75) Forgacs, E. Anal. Lett. 1994, 27(6), 1075-93. (H76) Forgacs, E. Anal. Chim. Acta 1994, 296(3), 235-41. (H77) Cserhati, T. Anal. Lett. 1994, 27(13), 2615-37. (H78) Cserhati, T.; Forgacs, E. Chemom. Intell. Lab. Syst. 1995, 28(2), 305-13. (H79) Olsen, B. A.; Sullivan, G. R. J. Chromatogr., A 1995, 692 (1 + 2), 147-59. (H80) Hamoir, T.; Sanchez, F. C.; Bourguignon, B.; Massart, D. L. J. Chromatogr. Sci. 1994, 32(11), 488-98. (H81) Altomare, C.; Carotti, A.; Cellamare, S.; Fanelli, F.; Gasparrini, F.; Villani, C.; Carrupt, P. A.; Testa, B. Chirality 1993, 5(7), 527-37. (H82) Nurok, D.; Kleyle, R. M.; Lipkowitz, K. B.; Myers, S. S.; Kearns, M. L. Anal. Chem. 1993, 65(24), 3701-7. (H83) Gonzalez, A. G.; Gonzalez-Arjona, D. Anal. Chim. Acta 1995, 312(3), 295-306. (H84) Altomare, C.; Cellamare, S.; Carotti, A.; Ferappi, M. Farmaco 1994, 49(6), 393-401. (H85) Park, J. H.; Chae, J. J.; Nah, T. H.; Jang, M. D. J. Chromatogr., A 1994, 664(2), 149-58. (H86) Poole, S. K.; Poole, C. F. J. Chromatogr., A 1995, 697(1 + 2), 429-40. (H87) Poole, S. K.; Poole, C. F. Analyst 1995, 120(2), 289-94. (H88) Poole, S. K.; Poole, C. F. J. Chromatogr., A 1995, 697(1 + 2), 415-27. (H89) Poole, S. K.; Kollie, T. O.; Poole, C. F. J. Chromatogr., A 1994, 664(2), 229-51. (H90) Clouser, D. L.; Jurs, P. C. Anal. Chim. Acta 1994, 295(3), 221-31. (H91) Svozil, D.; Pospichal, J.; Kvasnick, V. J. Chem. Inf. Comput. Sci. 1995, 35(5), 924-8. (H92) Balasubramanian, K. J. Chem. Inf. Comput. Sci. 1995, 35(2), 243-50. (H93) Xu, J.; Borer, P. N. J. Chem. Inf. Comput. Sci. 1994, 34(2), 349-56. (H94) van Geerestein-Ujah, E. C.; Slijper, M.; Boelens, R.; Kaptein, R. J. Biomol. NMR 1995, 6(1), 67-78. (H95) Novic, M.; Zupan, J. J. Chem. Inf. Comput. Sci. 1995, 35(3), 454-66. (H96) Stahnke, L. H. Meat Sci. 1995, 41(2), 211-23. (H97) Da Silva, M. A. A. P.; Elder, V.; Lederer, C. L.; Lundahl, D. S.; McDaniel, M. R. Dev. Food Sci. 1993, 33, 707-38. (H98) van Lill, D.; Wentzel, B. S.; Smith, M. F.; de Villiers, O. T. S. Afr. J. Plant Soil 1993, 10(4), 162-7. (H99) Vallejo-Cordoba, B.; Nakai, S. J. Agric. Food Chem. 1994, 42(4), 989-93. PATTERN RECOGNITION (I1) Otto, M. Spec. Publ.-R. Soc. Chem. 1994, 154 (Reviews on Analytical ChemistrysEuroanalysis VIII), 195-213. (I2) Smits, J. R. M.; Melssen, W. J.; Derksen, M. W. J.; Kateman, G. Anal. Chim. Acta 1993, 284(1), 91-105. (I3) Radomski, J. P.; van Halbeek, H.; Meyer, B. Nat. Struct. Biol. 1994, 1(4), 217-8. (I4) Meyer, M.; Meyer, K.; Hobert, H. Anal. Chim. Acta 1993, 282(2), 407-15. (I5) Li, Y.; van Espen, P. Chemom. Intell. Lab. Syst. 1994, 25(2), 241-8. (I6) Werther, W.; Lohninger, H.; Stanci, F.; Varmuza, K. Chemom. Intell. Lab. Syst. 1994, 22(1), 63-76. (I7) Tanabe, K.; Uesaka, H. Comput. Aided Innovation New Mater. 2; Doyama, M., Ed.; North-Holland: Amsterdam, The Netherlands, 1993; pp 943-8. (I8) Visser, T.; Luinge, H. J.; van der Maas, J. H. Anal. Chim. Acta 1994, 296(2), 141-54. (I9) Luinge, H. J.; van der Maas, J. H.; Visser, T. Chemom. Intell. Lab. Syst. 1995, 28(1), 129-38. (I10) Visser, T.; Luinge, H.-J.; van der Maas, J. Proc. SPIE-Int. Soc. Opt. Eng. 1993, 2089 (9th International Conference on Fourier Transform Spectroscopy), 232-3.
(I11) Sharma, A. K.; Sheikh, S.; Pelczer, I.; Levy, G. C. J. Chem. Inf. Comput. Sci. 1994, 34(5), 1130-9. (I12) Wienke, D.; Kateman, G. Chemom. Intell. Lab. Syst. 1994, 23(2), 309-29. (I13) Sato, T. J. Near Infrared Spectrosc. 1993, 1(4), 199-208. (I14) Schulze, H. G.; Blades, M. W.; Bree, A. V.; Gorzalka, B. B.; Greek, L. S.; Turner, R. F. B. Appl. Spectrosc. 1994, 48(1), 507. (I15) Hare, B. J.; Prestegard, J. H. J. Biomol. NMR 1994, 4(1), 3546. (I16) Dumitrescu, D.; Sarbu, C.; Pop, H. Anal. Lett. 1994, 27(5), 1031-54. (I17) Betti, A.; Lodi, G.; Fuzzati, N. J. Planar Chromatogr.-Mod. TLC 1993, 6(3), 232-7. (I18) Vervoort, R. J. M.; Derksen, M. W. J.; Maris, F. A. J. Chromatogr., A 1994, 678(1), 1-15. (I19) Juvancz, Z.; Cserhati, T.; Markides, K. E.; Bradshaw, J. S.; Lee, M. L. Chromatographia 1994, 38(3-4), 227-31. (I20) Haldna, U.; Pentchuk, J.; Righezza, M.; Chretien, J. R. J. Chromatogr., A 1994, 670(1-2), 51-8. (I21) Sanchez, F. C.; Lewi, P. J.; Massart, D. L. Chemom. Intell. Lab. Syst. 1994, 25(2), 157-77. (I22) Liang, Y.-z.; Haemaelaeinen, M. D.; Kvalheim, O. M.; Andersson, R. J. Chromatogr., A 1994, 662(1), 113-22. (I23) Goepel, W.; Schierbaum, K. D. Sens. Actuators, B 1995, B26(1-3), 1-12. (I24) Bull, D. R.; Harris, G. J.; Ben Rashed, A. B. Sens. Actuators, B 1993, 15(1-3), 151-61. (I25) Hashem, S.; Keller, R. E.; Kouzes, R. T.; Kangas, L. J. Proc. SPIE-Int. Soc. Opt. Eng. 1995, 2492 (Part 1), 33-40. (I26) Di Natale, C.; Davide, F.; D’Amico, A. Sens. Actuators, B 1995, B23(2-3), 111-8. (I27) Di Natale, C.; Davide, F. A. M.; D’Amico, A.; Hierlemann, A.; Mitrovics, J.; Schweizer, M.; Weimar, U.; Goepel, W. Sens. Actuators, B 1995, B25(1-3), 808-12. (I28) Di Natale, C.; Davide, F. A. M.; D’Amico, A. Sens. Actuators, B 1995, B27(1-3), 237-41. (I29) Stetter, J. R.; Findlay, M. W., Jr.; Schroeder, K. M.; Yue, C.; Penrose, W. R. Anal. Chim. Acta 1993, 248(1), 1-11. (I30) Rebiere, D.; Bordieu, C.; Piestre, J. Sens. Actuators, B 1995, B25(1-3), 777-80. (I31) Chang, S.-M.; Muramatsu, H.; Karube, I. Sens. Mater. 1995, 7(1), 13-22. (I32) Rapp, M.; Boss, B.; Voigt, A.; Gemmeke, H.; Ache, H. J. J. Anal. Chem. 1995, 352(7-8), 699-704. (I33) Ide, J.; Nakamoto, T.; Moriizumi, T. Sens. Actuators, B 1993, 13(1-3), 351-4. (I34) Auge, J.; Hauptmann, P.; Hartmann, J.; Roesler, S.; Lucklum, R. Sens. Actuators, B 1995, B26(1-3), 181-6. (I35) Niebling, G.; Schlachter, A. Sens. Actuators, B 1995, B27(13), 289-92. (I36) Adier, B.; Brueckner, G.; Winterstein, M. Chem. Tech. (Leipzig) 1994, 46(2), 77-86. (I37) Pisanelli, A. M.; Qutob, A. A.; Travers, P.; Szyszko, S.; Persaud, K. C. Life Chem. Rep. 1994, 11(2), 303-8. (I38) Gardner, J. W.; Pearce, T. C.; Friel, S.; Bartlett, P. N.; Blair, N. Sens. Actuators, B 1994, 18(1-3), 240-3. (I39) Di Natale, C.; Davide, F. A. M.; D’Amico, A.; Sberveglieri, G.; Nelli, P.; Faglia, G.; Perego, C. Sens. Actuators, B 1995, B25(13), 801-4. (I40) Nanto, H.; Tsubakino, S.; Ikeda, M.; Endo, F. Sens. Actuators, B 1995, B25(1-3), 794-6. (I41) Ramos, L. S. J. Chromatogr. Sci. 1994, 32(6), 219-27. (I42) Glickman, S. E.; Kilburn, J. O.; Butler, W. R.; Ramos, L. S. J. Clin. Microbiol. 1994, 32(3), 740-5. (I43) Garrity, G. M.; Heimbuch, B. K.; Motamedi, H.; Shafiee, A. J. Ind. Microbiol. 1993, 12(1), 42-7. (I44) Bouzar, H.; Jones, J. B.; Hodge, N. C. Phytopathology 1993, 83(7), 733-9. (I45) Romalde, J. L.; Magarinos, B.; Turnbull, K. D.; Baya, A. M.; Barja, J. M.; Toranzo, A. E. Arch. Microbiol. 1995, 163(3), 2116. (I46) Mergaert, J.; Wouters, A.; Swings, J. Syst. Appl. Microbiol. 1995, 17(4), 601-12. (I47) Jarvis, B. D. W.; Tighe, S. W. Plant Soil 1994, 161(1), 31-41. (I48) Yang, P.; Vauterin, L.; Vancanneyt, M.; Kersters, K.; Swings, J. Colloq. INRA 1994, 66, 265-70. (I49) Huys, G.; Vancanneyt, M.; Coopman, R.; Janssen, P.; Palsen, E.; Altwegg, M.; Kersters, K. Int. J. Syst. Bacteriol. 1994, 44(4), 651-8. (I50) Chum, J.; Atalan, E.; Kim, S. B.; Kim, H. J.; Hamid, M. E.; Trujillo, M. E.; Magee, J. G.; Manfio, G. P.; Ward, A. C.; Goodfellow, M. FEMS Microbiol. Lett. 1993, 114(1), 115-9. (I51) Sisson, P. R.; Freeman, R.; Law, D.; Ward, A. C.; Lightfoot, N. F. J. Anal. Appl. Pyrolysis, 1995, 32, 179-85. (I52) Basile, F.; Voorhees, K. J.; Hadfield, T. L. Appl. Environ. Microbiol. 1995, 61(4), 1534-9. (I53) Mavrovouniotis, M. L.; Harper, A. M.; Ifarraguerri, A. I. J. Chemom. 1994, 8(5), 305-31. (I54) Ge, Z.; Brown, C. W.; Kisner, H. J. Appl. Spectrosc. 1995, 49(4), 432-6. (I55) Andrasi, E.; Varga, I.; Dozsa, A.; Reffy, A.; Nagy, G. J. Chemom. Intell. Lab. Syst. 1994, 22(1), 107-14. (I56) Emard, J. F.; Andre, P.; Thouez, J. P.; Mathieu, J.; Boily, C.; Beaudry, M.; Cholette, A.; Robitaille, Y,; Bouchard, R.; et al. Water, Air, Soil, Pollut. 1994, 72(1-4), 251-64.
(I57) Choo, L.-P.; Mansfield, J. R.; Pizzi, N.; Somorjai, R. L.; Jackson, M.; Halliday, W. C.; Mantsch, H. H. Biospectroscopy 1995, 1(2), 141-8. (I58) Rietjens, M.; Wils, E. R. J. Propellants, Explos., Pyrotech. 1995, 20(4), 182-6. (I59) Fox, S.; Hooley, T. Proc. SPIE-Int. Soc. Opt. Eng. 1994, 2093 (Substance Identification Analytics), 195-203. (I60) Freeman, R.; Wheeler, J.; Sisson, P. R.; Ward, A. C.; Robertson, H.; Lightfoot, N. F. J. Anal. Appl. Pyrolysis 1994, 28(1), 3943. (I61) Jonson, C. S. L. Forensic Sci. Int. 1994, 69(1), 45-54. (I62) Hida, M.; Mitsui, T.; Minami, Y.; Fujimura, Y. J. Anal. Appl. Pyrolysis 1995, 32, 197-204. (I63) Hiraoka, Y. J. Forensic Sci. 1994, 39(6), 1381-92. (I64) Gao, N.; Cheng, M.-D.; Hopke, P. K. Atmos. Environ. 1994, 28(8), 1447-70. (I65) Mizohata, A.; Ito, N.; Masuda, Y. Taiki Kankyo Gakkaishi 1995, 30(4), 243-55. (I66) Michaud, D.; Baril, M.; Perrault, G. Air Waste 1993, 43(5), 729-35. (I67) Mendez, J.; Quejido, A. J.; Perez-Pastor, R.; Perez-Garcia, M. Anal. Chim. Acta 1993, 283(1), 528-37. (I68) Wienke, D.; Xie, Y.; Hopke, P. K. Chemom. Intell. Lab. Syst. 1994, 25(2), 367-87. (I69) Wienke, D.; Xie, Y.; Hopke, P. K. Anal. Chim. Acta 1995, 310(1), 1-14. (I70) Wienke, D.; Hopke, P. K. Anal. Chim. Acta 1994, 291(1-2), 1-18. (I71) Andrade, J. M.; Prada, D.; Alonso, E.; Lopez, P.; Muniategui, S.; de la Fuente, P.; Quijano, M. A. Anal. Chim. Acta 1994, 292(3), 253-61. (I72) Henrion, G.; Nass, D.; Michael, G.; Henrion, Fresenius’ J. Anal. Chem. 1995, 352(5), 431-6. (I73) Reisenhofer, E.; Picciotto, A.; Li, D. Anal. Chim. Acta 1995, 306(1), 99-106. (I74) Camarero, L.; Catalan, J.; Boggero, A.; Marchetto, A.; Mosello, R.; Psenner, R. Limnologica 1995, 25(2), 141-56. (I75) Kovats, Z. E.; Ciborowski, J. J. H. Environ. Monit. Assess. 1993, 27(2), 135-58. (I76) Burse, V. W.; Groce, D. F.; Caudill, S. P.; Korver, M. P.; Phillips, D. L.; McClure, P. C.; Lapeza, C. R., Jr.; Head, S. L.; Miller, D. T.; et al. Sci. Total Environ. 1994, 144(1-3), 153-77. (I77) Lavine, B. K.; Mayfield, H.; Kromann, P. R.; Faruque, A. Anal. Chem. 1995, 67(21), 3846-52. (I78) Andrews, J. M.; Lieberman, S. H. Anal. Chim. Acta 1994, 285(1-2), 237-46. (I79) Alaruri, S. D.; Rasas, M.; Alamedine, O.; Jubian, S.; Al-Bahrani, F.; Quinn, M. Opt. Eng. 1995, 34(1), 214-21. (I80) Nakai, S.; Aishima, T.; Yada, R. Y. In Protein Structure-Function Relationships in Foods; Yada, R. Y., Jackman, R. L., Smith, J. L., Eds.; Blackie: Glasgow, UK, 1994; pp 120-42. (I81) Matinez-Anaya, M. A. Rev. Esp. Cienc. Tecnol. Aliment. 1994, 34(5), 469-93. (I82) Barroso, C. G.; Garcia, M. V.; Perez-Bustamante, J. A. Bull. Liaison-Groupe Polyphenols 1992, 16 (Part 2), 64-8. (I83) Palma, M.; Barroso, C. G.; Perez-Bustamante, J. A. Quim. Anal. (Barcelona) 1994, 13(2), 102-5. (I84) Garcia-Jares, C.; Garcia-Martin, S.; Cela-Torrijos, R. J. Agric. Food Chem. 1995, 43(3), 764-8. (I85) Shimoda, M.; Shibamoto, T.; Noble, A. C. J. Agric. Food Chem. 1993, 41(10), 1664-8. (I86) Moio, L.; Schlich, P.; Etievant, P. Sci. Aliments 1994, 14(5), 601-8. (I87) Garcia-Jares, C. M.; Garcia-Martin, M. S.; Carro-Marino, N.; Cela- Torrijos, R. J. Sci. Food Agric. 1995, 69(2), 175-84. (I88) Moret, I.; Scarponi, G.; Cescon, P. J. Agric. Food Chem. 1994, 42(5), 1143-53. (I89) Latorre, M. J.; Garcia-Jares, C.; Medina, B.; Herrero, C. J. Agric. Food Chem. 1994, 42(7), 1451-5. (I90) Servili, M.; Conner, J. M.; Piggot, J. R.; Withers, S. J.; Paterson, A. J. Sci. Food Agric. 1995, 67(1), 61-70. (I91) Morales, M. T.; Alonso, M. V.; Rios, J. J.; Aparicio, R. J. Agric. Food Chem. 1995, 43(11), 2925-31. (I92) Wesley, I. J.; Barnes, R. J.; McGill, A. E. J. J. Am. Oil Chem. Soc. 1995, 72(3), 289-92. (I93) Drava, G.; Forina, M.; Lanteri, S.; Lupoli, M. J. Sci. Food Agric. 1994, 65(1), 21-30. (I94) Tsimidou, M.; Karakostas, K. X. J. Sci. Food Agric. 1993, 62(3), 253-7. (I95) Campisi, B.; Dugo, G.; Cotroneo, A.; Favretto, L. Anal. Chim. Acta 1995, 312(2), 199-205. (I96) Sato, T. J. Am. Oil Chem. Soc. 1994, 71(3), 293-8. (I97) Maffei, M.; Mucciarelli, M.; Scannerini, S. Biochem. Syst. Ecol. 1994, 22(7), 679-87. (I98) Chialva, F.; Ariozzi, A.; Decastri, D.; Manitto, P.; Clementi, S.; Bonelli, D. J. Agric. Food Chem. 1993, 41(11), 2028-33. (I99) Lamberto, M.; Saitta, M. J. Am. Oil Chem. Soc. 1995, 72(8), 867-71. (I100) Lai, Y. W.; Kemsley, E. K.; Wilson, R. H. J. Agric. Food Chem. 1994, 42(5), 1154-9. (I101) Safar, M.; Bertrand, D.; Robert, P.; Devaux, M. F.; Genot, C. J. Am. Oil Chem. Soc. 1994, 71(4), 371-7. (I102) Jurasek, P.; Kosik, M.; Phillips, G. O.; Varmuza, K. Foods Food Ingredients J. 1993, 157, 78-91. (I103) Jurasek, P.; Kosik, M.; Phillips, G. O. Food Hydrocolloids 1993, 7(2), 157-74.
Analytical Chemistry, Vol. 68, No. 12, June 15, 1996
(I104) Jurasek, P.; Kosik, M.; Phillips, G. O. Food Hydrocolloids 1993, 7(3), 255-80. (I105) Jurasek, P.; Phillips, G. O.; Varga, S.; Chikamai, B. N.; Banks, W. B. Food Hydrocolloids 1994, 8(6), 567-88. (I106) Biswas, S.; Biswas, B.; Phillips, G. O. Food Hydrocolloids 1995, 9(3), 151-63. (I107) Kemsley, E. K.; Ruault, S.; Wilson, R. H. Food Chem. 1995, 54(3), 321-6. (I108) Dupuy, N.; Huvenne, J. P.; Duponchel, L.; Legrand, P. Appl. Spectrosc. 1995, 49(5), 580-5. (I109) Murota, A. Biosci., Biotechnol., Biochem. 1993, 57(7), 10438. (I110) Bicchi, C. P.; Binello, A. E.; Pellegrino, G. M.; Vanni, A. C. J. Agric. Food Chem. 1995, 43(6), 1549-55. (I111) Tomlins, K. I.; Gay, C. Food Chem. 1994, 50(2), 157-65. (I112) Ulberth, F. J. AOAC Int. 1994, 77(5), 1326-34. (I113) Vallejo-Cordoba, B.; Nakai, S. J. Agric. Food Chem. 1994, 42(4), 994-9. (I114) Defernez, M.; Wilson, R. H. J. Sci. Food Agric. 1995, 67(4), 461-7. (I115) Defernez, M.; Kemsley, E. K.; Wilson, R. H. J. Agric. Food Chem. 1995, 43(1), 109-13. (I116) Robards, K.; Antolovich, M. Chem. Aust. 1994, 61(7), 392-5. (I117) Twomey, M.; Downey, G.; McNulty, P. B. J. Sci. Food Agric. 1995, 67(1), 77-84. (I118) Shaw, P. E.; Moshonas, M. G.; Buslig, B. S. ACS Symp. Ser. 1995, No. 596 (Fruit Flavors), 33-47. (I119) Evans, D. G.; Scotter, C. N. G.; Day, L. Z.; Hall, M. N. J. Near Infrared Spectrosc. 1993, 1(1), 33-44. (I120) van den Broek, W. H. A. M.; Wienke, D.; Melssen, W. J.; de Crom, C. W. A.; Buydens, L. Anal. Chem. 1995, 67(20), 37539. (I121) Scott, D. M.; Waterland, R. L. Polym. Eng. Sci. 1995, 35(12), 1011-5. (I122) Alam, M. K.; Stanton, S. L.; Hebner, G. A. Spectroscopy (Eugene, Oreg.) 1994, 9(2), 30, 32-4, 36-8, 40. (I123) Jansen, J.; Hastenberg, C. v. Kunststoffe 1994, 84(1), 51-4. (I124) Gilbert, C.; Kokot, S. Vib. Spectrosc. 1995, 9(2), 161-7. (I125) Kokot, S.; Stewart, S. Text. Res. J. 1995, 65(11), 643-51. (I126) Kokot, S.; Gilbert, C. Analyst (Cambridge, U.K.) 1994, 119(4), 671-6. (I127) Downey, G. Analyst (Cambridge, U.K.) 1994, 119(11), 236775. (I128) Cuesta Sanchez, F.; Toft, J.; van den Bogaert, B.; Massart, D. L.; Dive, S. S.; Hailey, P. Fresenius’ J. Anal. Chem. 1995, 352(7-8), 771-8. (I129) Dempster, M. A.; MacDonald, B. F.; Gemperline, P. J.; Boyer, N. R. Anal. Chim. Acta 1995, 310(1), 43-51. (I130) Carlsson, A. E.; Janne, K. L.-R. Appl. Spectrosc. 1995, 49(7), 1037-40. (I131) Gemperline, P. J.; Boyer, N. R. Anal. Chem. 1995, 67(1), 1606. (I132) Ricker, R. D.; Sandoval, L. A.; Justice, J. D.; Geiser, F. O. J. Chromatogr., A 1995, 691(1-2), 67-79. (I133) Hartmann, C.; Massart, D. L. J. AOAC Int. 1994, 77(5), 131825. (I134) Cho, J.; Gemperline, P. J. J. Chemom. 1995, 9(3), 169-78. (I135) Nierop, A. F. M.; Tas, A. C.; van der Greef, J. Chemom. Intell. Lab. Syst. 1994, 25(2), 249-63. (I136) Wienke, D.; Buydens, L. Trends Anal. Chem. 1995, 14(8), 398406. (I137) Forina, M.; Lanteri, S.; Boggia, R.; Bertran, E. Quim. Anal. (Barcelona) 1993, 12(3), 128-35. LIBRARY SEARCHING (J1) Dathe, B.; Otto, M. Chromatographia, 1993, 37(1-2), 31-6. (J2) Brown, C.; Okafor, A.; Donahue, S.; Su-Chin, L. Appl. Spectrosc. 1995, 49(7), 1022-7. (J3) Klawun, C.; Wilkins, C. L. Anal. Chem., 1995, 67(2), 374-78. (J4) Mittermayr, C. R.; Drouen, A. C. J. H.; Otto, M.; Grasserbauer, M. Anal. Chim. Acta., 1994, 294(3), 227-42. (J5) Chen, L.; Robien, W. J. Chem. Inf. Comput. Sci., 1994, 34(4), 934-41. (J6) Gross, A. R.; Adams, M. J. Anal. Proc. 1994, 3(1), 22-5. (J7) Stein, S. E. J. Am. Soc. Mass. Spectrom. 1994, 5(4), 316-23. (J8) Stein, S. E. J. Am. Soc. Mass. Spectrom. 1995, 6(8), 644-55. (J9) Stein, S. E.; Scott, D. R. J. Am. Soc. Mass. Spectrom. 1994, 5(9), 859-66. (J10) Boruta, M.; Carl, R. T.; Grim, W. M.; Tontala, F. J. Proc. SPIEInt. Soc. Opt. Eng. 1993, 2089, 328-9. (J11) Durindova, M.; Jurasek, P.; Kibik, I.; Kovacik, V. Chem. Pap. 1993, 47(3), 198-201. (J12) Smith, N. B. J. Anal. Toxicol. 1994, 18(1), 16-21. (J13) Stan, H.-J.; Schwarzer, F. J. Chromatogr. 1993, 653(1), 4553. (J14) Postma, G. J.; Hack, F. M.; Janssen, P. A.; Buydens, L. M. C.; Kateman, G. Chemom. Intell. Lab. Syst. 1994, 25(2), 285-95. (J15) Ojanperae, I.; Jaenchen, P. In Contributions in Forensic Toxicology; Mueller, R. K., Eds.; MOLINApress: Leipzig, Germany, 1994; pp 394-8. (J16) Thomas, R. G.; Watkinsion, D. E. Stud. Conserv. 1994, 39(3), 187-92. (J17) Dempster, M. A.; Jones, J. A.; Last, I. R.; MacDonald, B. F.; Prebble, K. A. J. Pharm. Biomed. Anal. 1993, 11(11-2), 108792. 60R
Analytical Chemistry, Vol. 68, No. 12, June 15, 1996
(J18) Yates, J. R.; Eng, J. K.; McCormack, A. L.; Schieltz, D. Anal. Chem. 1995, 67(8), 1426-36. (J19) Laber, H.; Schulz, J.; Sponholz, W. R.; Bremser, W. Fresnius J. Anal. Chem. 1995, 351(6), 530-5. (J20) Rozenblum, A.; Brunerie, P. Dev. Food. Sci. 1994, 35, 133-7. (J21) Schuberth, J. J. Chromatogr., A 1994, 674(1-2), 63-71. (J22) Fuller, M.; Rosenthal, R. Proc. SPIE-Int. Soc. Opt. Eng. 1993, 2089, 440-1. ARTIFICIAL INTELLIGENCE (K1) Bryant, C. H.; Adam, A.; Taylor, D. R.; Rowe, R. C. Anal. Chim. Acta 1994, 297, 317-47. (K2) Bell, S. E.; Mead, W. C. In Computer Enhanced Analytical Spectroscopy; Wilkins, C. L., Ed.; Plenum: New York, 1993; Vol. 4, pp 275-303. (K3) Barton, G. L. Curr. Opin. Struct. Biol. 1995, 5, 372-6. (K4) Sapegin, A. M. In Trends in QSAR and Molecular Modelling 92, Proc. Eur. Symp. Struct.-Act. Relat.; Wermuth, C. G., Ed.; ESCOM: Leiden, Netherlands, 1993; pp 495-6. (K5) Rius, F. X. Anal. Chim. Acta 1993, 283, 518-27. (K6) Leser, E. W.; Asenjo, J. A. Ann. N.Y. Acad. Sci. 1994, 721, 337-47. (K7) Place, J. F.; Truchaud, A.; Ozawa, K.; Pardue, H.; Schnipelsky, P. J. Autom. Chem. 1995, 17, 1-15. (K8) Lucasius, C. B.; Kateman, G. Chemom. Intell. Lab. Syst. 1994, 25, 99-145. (K9) Davis, M. J. Adv. Ser. Phys. Chem. 1995, 4, 855-90. (K10) Wagner, J. S.; Trahan, M. W.; Nelson, W. E.; Hargis, P. J., Jr.; Tisone, G. C. Proc. SPIE-Int. Soc. Opt. Eng. 1995, 2367, 22738. (K11) Fekete, J.; Morovjan, G.; Csizmadia, F.; Darvas, F. J. Chromatogr., A 1994, 660, 33-46. (K12) Wehrens, R.; Buydens, L. M. C.; Hindriks, H.; Maris, F. A. Chemom. Intell. Lab. Syst. 1994, 25, 341-54. (K13) Hubert, P.; Chiap, P.; Moors, M.; Bourguignon, B.; Massart, D. L.; Crommen, J. J. Chromatogr., A 1994, 665, 87-99. (K14) Mulholland, M.; Hibbert, D. B.; Haddad, P. R.; Sammut, C. Chemom. Intell. Lab. Syst. 1995, 27, 95-104. (K15) Gros, N.; Gorenc, B. J. Chromatogr., A 1995, 697, 31-43. (K16) Pingand, P. B.; Lerner, D. A. Anal. Chim. Acta 1994, 293, 301-9. (K17) Penninckx, W.; Smeyers-Verbeke, J.; Massart, D. L.; Spanjers, L. G. C. W.; Maris, F. A. Anal. Chim. Acta 1993, 282, 41722. (K18) Esteban, M.; Arino, C.; Ruisanchez, I.; Larrechi, M. S.; Rius, F. X. Anal. Chim. Acta 1993, 284, 435-43. (K19) Esteban, M.; Arino, C.; Ruisanchez, I.; Larrechi, M. S.; Rius, F. X. Anal. Chim. Acta 1994, 285, 193-208. (K20) Esteban, M.; Arino, C.; Ruisanchez, I.; Larrechi, M. S.; Rius, F. X. Anal. Chim. Acta 1994, 285, 377-89. (K21) Du, H.; Lahiri, S.; Huang, G.; Stillman, M. J. Anal. Chim. Acta 1994, 296, 21-31. (K22) Du, H.; Stillman, M. J. Anal. Chim. Acta 1994, 296, 33-41. (K23) Lahiri, S.; Yuan, B.; Stillman, M. J. Anal. Chem. 1994, 66, 2954-63. (K24) Webb, D. J.; Hamier, J.; Salin, E. D. Trends Anal. Chem. 1994, 13, 44-53. (K25) Workman, J., Jr.; Kowalski, B.; Mobley, P. Proc. Annu. ISA Anal. Div. Symp. 1995, 28, 97-106. (K26) Brandt, J.; Hitzmann, B. Anal. Chim. Acta 1994, 291, 29-40. (K27) Hitzmann, B.; Lohn, A.; Reinecke, M.; Schulze, B.; Scheper, T. Anal. Chim. Acta 1995, 313, 55-62. (K28) Funatsu, K.; Nishizaki, M.; Sasaki, S. J. Chem. Inf. Comput. Sci. 1994, 34, 745-51. (K29) Peng, C.; Yuan, S.; Zheng, C.; Hui, Y.; Wu, H.; Ma, K.; Han, X. J. Chem. Inf. Comput. Sci. 1994, 34, 814-9. (K30) Zimmerman, D.; Kulikowski, C.; Wang, L.; Lyons, B.; Montelione, G. T. J. Biomol. NMR 1994, 4, 241-56. (K31) Aarnio, P. A.; Ala-Heikkila, J. J.; Hakulinen, T. T.; Routti, J. T. J. Radioanal. Nucl. Chem. 1995, 193, 219-27. (K32) Andreev, G. N.; Argirov, O. K.; Penchev, P. N. Anal. Chim. Acta 1993, 284, 131-6. (K33) Andreev, G. N.; Argirov, O. K. J. Mol. Struct. 1995, 347, 43948. (K34) Moore, D. S.; White, J. S.; Harbin, B. A. Anal. Chim. Acta 1994, 294, 85-94. (K35) Huixiao, H.; Xinquan, X. J. Chem. Inf. Comput. Sci. 1994, 34, 730-4. (K36) Huixiao, H.; Xinquan, X. J. Chem. Inf. Comput. Sci. 1994, 34, 1259-66. (K37) Huixiao, H.; Yinling, H.; Xinquan, X.; Yufeng, S. J. Chem. Inf. Comput. Sci. 1995, 35, 979-1000. (K38) Elling, J. W.; Mniszewski, S. M.; Zahrt, J. D. J. Chromatogr. Sci. 1994, 32, 213-8. (K39) Scott, D. R. Chemom. Intell. Lab. Syst. 1994, 23, 351-64. (K40) Schechter, I.; Wisbrun, R.; Niessner, R.; Schroeder, H.; Kompa, K. L. Proc. SPIE-Int. Soc. Opt. Eng. 1994, 2092, 174-85. (K41) Hu, C.; Xu, L. Anal. Chim. Acta 1994, 298, 75-85. (K42) Hu, C.; Xu, L. Sci. China, Ser. B 1995, 38, 296-304. (K43) Feinberg, M.; Suard, C.; Ireland-Ripert, J. Chemom. Intell. Lab. Syst. 1994, 22, 37-47. (K44) Rose, J. R.; Gasteiger, J. J. Chem. Inf. Comput. Sci. 1994, 34, 74-90. (K45) Zhu, Q.; Stillman, M. J. J. Chem. Inf. Comput. Sci. 1995, 35, 945-55.
(K46) Zhu, Q.; Stillman, M. J. J. Chem. Inf. Comput. Sci. 1995, 35, 956-68. (K47) Tetko, I. V.; Livingstone, D. J.; Luik, A. I. J. Chem. Inf. Comput. Sci. 1995, 35, 826-33. (K48) Smits, J. R. M.; Melssen, W. J.; Buydens, L. M. C.; Kateman, G. Chemom. Intell. Lab. Syst. 1994, 22, 165-89. (K49) Smits, J. R. M.; Melssen, W. J.; Buydens, L. M. C.; Kateman, G. Chemom. Intell. Lab. Syst. 1994, 23, 267-91. (K50) Burden, F. R. J. Chem. Inf. Comput. Sci. 1994, 34, 1229-31. (K51) Lu, T.; Lin, F.; Kostrzewski, A.; Lerner, J. Proc. SPIE-Int. Soc. Opt. Eng. 1994, 2093, 390-9. (K52) Klawun, C.; Wilkins, C. L. J. Chem. Inf. Comput. Sci. 1994, 34, 984-93. (K53) Blank, T. B., Brown, S. D. J. Chemom. 1994, 8, 391-408. (K54) Harrington, P. de B. Anal. Chem. 1994, 66, 802-7. (K55) Bro, R. J. Chemom. 1995, 9, 413-30. (K56) Vetter, K.; Elze, T. W.; Kaspar, M.; Kroell, T.; Gerl, J. Nucl. Instrum. Methods Phys. Res., Sect. A 1994, 344, 607-10. (K57) Reyes, M. V.; Pyeatt, L. D. U.S. Patent 5424959, June 13, 1995. (K58) Hussain, A. S.; Shivanand, P.; Johnson, R. D. Drug Dev. Ind. Pharm. 1994, 20, 1739-52. (K59) Bangov, I. P.; Laude, I.; Cabrol-Bass, D. Anal. Chim. Acta 1994, 298, 33-52. (K60) Otto, M. Anal. Chim. Acta 1993, 283, 500-7. (K61) de Agapito, J. A.; de Agapito, L.; Schneider, M.; Garcia Rosa, R.; de Pedro, T. Sens. Actuators, B 1993, 15, 105-9.
(K62) Arnold, T.; Otto, M.; Wegscheider, W. Talanta 1994, 41, 1169-84. (K63) Wright, D.; Liu, C. L.; Stanley, D.; Chen, H. C.; Fang, J. H. Comput. Geosci. 1993, 19, 1429-43. (K64) Lindsey, J. S.; Corkan, L. A. Chemom. Intell. Lab. Syst. 1993, 21, 139-50. (K65) Gentsch, J. Chemom. Intell. Lab. Syst. 1993, 21, 229-33. (K66) Garcia-Mesa, J. A.; Luque de Castro, M. D.; Valcarcel, M. Anal. Chem. 1993, 65, 3540-2. (K67) Levy, G. C.; Pelczer, I.; Jeong, G.; Wang, K. J. Magn. Reson., Ser. A 1993, 105, 316-20. (K68) Wieling, J.; Jongman, J. H. G.; Hempenius, J.; Mensink, C. K. Chemom. Intell. Lab. Syst. 1994, 25, 355-66. (K69) Mulholland, M.; Hibbert, D. B.; Haddad, P. R.; Parslov, P. Chemom. Intell. Lab. Syst. 1995, 30, 117-28. (K70) Tandler, P. J.; Butcher, J. A.; Tao, H.; Harrington, P. de B. Anal. Chim. Acta 1995, 312, 231-44. (K71) Liebman, S. A.; Phillips, C.; Fitzgerald, W.; Levy, E. J. ASTM Spec. Tech. Publ., STP 1994, 1221, 33-7. (K72) Schubert, J.; Simutis, R.; Dors, M.; Havlik, I.; Luebbert, A. J. Biotechnol. 1994, 35, 51-68. (K73) Walczak, B.; Bauer-Wolf, E.; Wegscheider, W. Mikrochim. Acta 1994, 113, 153-69.
Analytical Chemistry, Vol. 68, No. 12, June 15, 1996
Analytical Chemistry, Vol. 68, No. 12, June 15, 1996