Exploratory Multivariate Data Analysis

Cover The words on the front and back cover pages have been ordered by letting MATLAB ® choose randomly and uniformly from a list of selected keywords. i Preface This PhD dissertation is based on a series of research projects conducted at The Royal Veterinary and Agricultural University (KVL), in the Chemometrics Research Group of Professor Lars Munck and colleagues during the period 1996-99. Parts of the theoretical research were accomplished during a part-time position as guest researcher at Humboldt I am greatly indebted to the many skilled and visionary people with whom I have had the great privilege to collaborate. The very competent chemometrics and spectroscopic groups at Food Technology are gratefully thanked for offering numerous interesting challenges. First and foremost, my colleagues in the group of Food Technology and in particular Lars Munck who through numerous discussions on the importance of context in relation to the exploratory spirit elevated my scope to look beyond the limits of the many isolated projects to find the " principal components " of successful solutions to problems based on data from systems of high complexity. With regards to algorithms and models, much appreciated professional and personal inspiration has come from Rasmus Bro and Lars Nørgaard. Søren B. Engelsen introduced me to the challenges of molecular modelling during a joint research project. Appreciation goes to Gilda Kischinovsky for proofreading of the text. thanked for a very educational stay in Berlin and for the close collaboration. Danisco Sugar Development Center, Nakskov, is thanked for providing industrial samples and data that have allowed for exploratory application of multivariate data analysis and, thus, many of the results reported in this dissertation. Salaries, computers and travels have been funded by The Nordic Industrial Fund and the European Commission projects AFFLUENCE (CT 94-1416) and NWAYQUAL (GRD1-1999-10337). The funds are much appreciated for the required financial support during the period. ii The thesis is based on work published in the following peer-reviewed papers which are included as off-prints at the back of the thesis and are referred to throughout the text by labels P1-P10: P2 A new criterion for simple-structure transformations of core arrays in N-way principal components analysis, R. P7 Chemometrics in food science-a demonstration of the feasibility of a highly exploratory, inductive evaluation strategy of fundamental scientific significance, L. P8 Multi-way chemometrics for mathematical separation of fluorescent colorants and colour precursors from spectrofluorimetry of beet sugar and …