Methodological contribution to control heteroscedasticity in discriminant analysis studies

Abstract

We describe for two groups, the process of establishing an heteroscedastic model in Monte Carlo discriminant analysis studies. The simple model proposed allows, by the linear transformation, to extend the results of discriminant analysis studies to a large variety of situations. The heteroscedasticity degree of the model is appreciated by a parameter defined in the study, which can be computed not only on populations but also on data samples. This model can then be used to express the results of Monte Carlo discriminant analysis studies as a function of the heteroscedasticity degree observed on data samples.

Keywords: heteroscedasticity, discriminant analysis, two groups, Monte Carlo studies

Global Journal of Pure and Applied Sciences Vol. 12(1) 2006: 107-110