Abstract
Regression analysis is the most commonly used technique in QSAR studies. Whilst regression has its advantages, it also suffers from a number of disadvantages. A variety of mathematical and statistical techniques exist which complement or replace regression. These may he classified as ‘supervised’ and ‘unsupervised’ learning methods where supervision refers to the use made of biological data in the analysis. These terms are discussed and examples of three unsupervised and two supervised learning methods are presented. The techniques described are Non‐Linear Mapping, Principal Components Analysis, Cluster Analysis, Canonical Correlation Analysis and Factor Analysis. Although the examples presented are mostly pharmaceutical applications, it is proposed that the extra complexity of pesticide research data, compared with pharmaceutical, makes it very well suited to analysis by these methods.

This publication has 27 references indexed in Scilit: