Discriminant Analysis

Abstract

In this article, we consider discriminant analysis where classification is undertaken in the context where there are a number of classes and the observed data (the so‐called training data) are of known classification with respect to these classes. We consider both parametric and nonparametric approaches to the formation of allocation rules from the available training data. The coverage starts with Fisher's linear discriminant function, which was the initial attempt to perform discriminant analysis in its modern guise. The estimation of the error rates of the sample‐based rules so formed are considered. Particular attention is given to the classification of high‐dimensional data where the number of training observations is small relative to the large number of available variables.

Behavioural Aspects of Office Space Structures in the City: The Case of Warsaw’s Business Districts

“…To identify differences between the two districts, a few other methods were adopted, depending on the character of the variables (dichotomous, ordinal, continuous): Chi-square test, Mann-Whitney test, and t-test of independence. Advanced linear discriminatory analysis (McLachlan, 2004) was used to identify the key factors that differentiated the CBD from the SBD. The data used in the discriminant analysis met assumptions specified for this method (Tabachnick & Fidell, 2019).…”