Multivariate Calibration Harald Martens, Chemist, Norwegian Food Research Institute, Aas, Norway and Norwegian Computing Center, Oslo, Norway Tormod Naes, Statistician, Norwegian Food Research Institute, Aas, Norway The aim of this inter-disciplinary book is to present an up-to-date view of multivariate calibration of analytical instruments, for use in research, development and routine laboratory and process operation. The book is intended to show practitioners in chemistry and technology how to extract the quantitative and understandable information embedded in non-selective, overwhelming and apparently useless measurements by multivariate data analysis. Multivariate calibration is the process of learning how to combine data from several channels, in order to overcome selectivity problems, gain new insight and allow automatic outlier detection. Multivariate calibration is the basis for the present success of high-speed Near-Infrared (NIR) diffuse spectroscopy of intact samples. But the technique is very general: it has shown similar advantages in, for instance, UV, Vis, and IR spectrophotometry, (transmittance, reflectance and fluorescence), for x-ray diffraction, NMR, MS, thermal analysis, chromatography (GC, HPLC) and for electrophoresis and image analysis (tomography, microscopy), as well as other techniques. The book is written at two levels: the main level is structured as a tutorial on the practical use of multivariate calibration techniques. It is intended for university courses and self-study for chemists and technologists, giving one complete and versatile approach, based mainly on data compression methodology in self-modelling PLS regression, with considerations of experimental design, data pre-processing and model validation. A second, more methodological, level is intended for statisticians and specialists in chemometrics. It compares several alternative calibration methods, validation approaches and ways to optimize the models. The book also outlines some cognitive changes needed in analytical chemistry, and suggests ways to overcome some communication problems between statistics and chemistry and technology.
This book offers an introductory-level guide to the complex field of multivariate analytical calibration, with particular emphasis on real applications such as near infrared spectroscopy. It presents intuitive descriptions of mathematical and statistical concepts, illustrated with a wealth of figures and diagrams, and consistently highlights physicochemical interpretation rather than mathematical issues. In addition, it describes an easy-to-use and freely available graphical interface, together with a variety of appropriate examples and exercises. Lastly, it discusses recent advances in the field (figures of merit, detection limit, non-linear calibration, method comparison), together with modern literature references.
"Multivariate Data Analysis - in practice adopts a practical, non-mathematical approach to multivariate data analysis. The book's principal objective is to provide a conceptual framework for multivariate data analysis techniques, enabling the reader to apply these in his or her own field. Features: Focuses on the practical application of multivariate techniques such as PCA, PCR and PLS and experimental design. Non-mathematical approach - ideal for analysts with little or no background in statistics. Step by step introduction of new concepts and techniques promotes ease of learning. Theory supported by hands-on exercises based on real-world data. A full training copy of The Unscrambler (for Windows 95, Windows NT 3.51 or later versions) including data sets for the exercises is available. Tutorial exercises based on data from real-world applications are used throughout the book to illustrate the use of the techniques introduced, providing the reader with a working knowledge of modern multivariate data analysis and experimental design. All exercises use The Unscrambler, a de facto industry standard for multivariate data analysis software packages. Multivariate Data Analysis in Practice is an excellent self-study text for scientists, chemists and engineers from all disciplines (non-statisticians) wishing to exploit the power of practical multivariate methods. It is very suitable for teaching purposes at the introductory level, and it can always be supplemented with higher level theoretical literature."Résumé de l'éditeur.
Using formal descriptions, graphical illustrations, practical examples, and R software tools, Introduction to Multivariate Statistical Analysis in Chemometrics presents simple yet thorough explanations of the most important multivariate statistical methods for analyzing chemical data. It includes discussions of various statistical methods, such as principal component analysis, regression analysis, classification methods, and clustering. Written by a chemometrician and a statistician, the book reflects the practical approach of chemometrics and the more formally oriented one of statistics. To enable a better understanding of the statistical methods, the authors apply them to real data examples from chemistry. They also examine results of the different methods, comparing traditional approaches with their robust counterparts. In addition, the authors use the freely available R package to implement methods, encouraging readers to go through the examples and adapt the procedures to their own problems. Focusing on the practicality of the methods and the validity of the results, this book offers concise mathematical descriptions of many multivariate methods and employs graphical schemes to visualize key concepts. It effectively imparts a basic understanding of how to apply statistical methods to multivariate scientific data.
This book supplies a unifying framework for the derivation of probabilistic membership values in any classification task. While statistical classifiers usually generate such probabilities which reflect the assessment uncertainty, regularization methods supply membership values which do not induce any probabilistic confidence. It is desirable, to transform or re-scale membership values to membership probabilities, since they are comparable and easier applicable for post-processing. In this book several univariate calibration methods are presented. The methods are compared by their performance in experiments measured in terms of correctness and well-calibration. Multivariate extensions for regularization methods usually use a reduction to binary tasks, followed by univariate calibration and further application of the pairwise coupling algorithm. This book introduces a well-performing alternative to coupling that generates Dirichlet distributed membership probabilities. This flexible one-step algorithm bases on probability theory and is applicable to all classification problems. Dirichlet calibration method and pairwise coupling are compared in further experiments.
This book presents new developments in data analysis, classification and multivariate statistics, and in their algorithmic implementation. The volume offers contributions to the theory of clustering and discrimination, multidimensional data analysis, data mining, and robust statistics with a special emphasis on the novel Forward Search approach. Many papers provide significant insight in a wide range of fields of application. Customer satisfaction and service evaluation are two examples of such emerging fields.
Practical Three-Way Calibration is an introductory-level guide to the complex field of analytical calibration with three-way instrumental data. With minimal use of mathematical/statistical expressions, it walks the reader through the analytical methodologies with helpful images and step-by-step explanations. Unlike other books on the subject, there is no need for prior programming experience and no need to learn programming languages. Easy-to-use graphical interfaces and intuitive descriptions of mathematical and statistical concepts make three-way calibration methodologies accessible to analytical chemists and scientists in a wide range of disciplines in industry and academia. Numerous detailed examples of slowly increasing complexity Exposure to several different data sets and techniques through figures and diagrams Computer program screenshots for easy learning without prior knowledge of programming languages Minimal use of mathematical/statistical expressions
Pattern recognition and other chemometrical techniques are important tools in interpreting environmental data. This volume presents authoritatively state-of-the-art applications of measuring and handling environmental data. The chapters are written by leading experts.