New approaches using chemometric methods: Categorical regression analysis and comparison of maximum-minimum distance clustering with other methods.
Item
-
Title
-
New approaches using chemometric methods: Categorical regression analysis and comparison of maximum-minimum distance clustering with other methods.
-
Identifier
-
AAI9315514
-
identifier
-
9315514
-
Creator
-
Wu, Jinan.
-
Contributor
-
Adviser: Darryl G. Howery
-
Date
-
1993
-
Language
-
English
-
Publisher
-
City University of New York.
-
Subject
-
Chemistry, Analytical | Chemistry, Physical | Statistics
-
Abstract
-
Two new chemometric methodologies, categorical regression analysis and maximum-minimum distance cluster analysis, are applied to a number of data sets from chemistry. Computer programs with documentation are written for both methods.;Independent variables in the models for categorical regression analysis (CRA) consist of 1's (substance has the property) and 0's (substance lacks the property) only. Comparing input data with calculated values, reasonable (and in some problems excellent) models were developed for a variety of data including retention indices, boiling points, dissociation constants and rate constants. The simplicity of CRA models is a major advantage of the approach. Applications are limited to problems in which the substances are structurally similar.;MRA-CRA combination models for retention-index problems are developed. MRA-CRA models combine the advantage of flexibility and increased statistical significance.;Assignment of data vectors to a selected number of clusters is accomplished using maximum-minimum (max-min) distance comparisons. Clusters from the max-min method are compared to clusters from three other approaches: hierarchical cluster analysis, varimax-rotated factor analysis and target factor analysis. Several GLC retention-index matrices and TLC R{dollar}\sb{lcub}\rm f{rcub}{dollar}-value matrices were studied. Nearly equivalent clusters are obtained using the various methods for the simpler GLC problems. For the more complicated GC problems and especially for the TLC problem (which involved mixed solvents), confusingly different clusters are obtained from the different methods.
-
Type
-
dissertation
-
Source
-
PQT Legacy CUNY.xlsx
-
degree
-
Ph.D.
