Chemometric investigations with minimal suppositions
Date
2018
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
University of Delaware
Abstract
Modern chemometric modeling methods tend to be extensions of base methods, and because of this they tend to impose additional assumptions about data in the pursuit of analysis. In many circumstances, very little is actually known about a set of data, and such assumptions are unfounded or can only be applied because of post analysis rationalizations. This thesis addresses several cases where there is very little a priori information available about the datasets and demonstrates new ways to extract qualitative or quantitative information. In this thesis three novel analytical methods are introduced and demonstrated in three areas of chemometrics. The three methods included in this work are: band target entropy minimization and target partial least squares for one at a time multivariate curve resolution, small moving window calibration models for soft sensing processes with limited history, and uncharted forest an exploratory data analysis method.
Description
Keywords
Pure sciences, Chemometric investigations, Minimal suppositions