A comparison of multivariate discriminant and logistic regression analysis in the prediction of corporate failure in South Africa

Abstract

The prediction of corporate failure, based on a specific model, is of vital importance to a large body of individuals. The two statistical methodologies which enjoy prominence in the literature when establishing these models are multivariate discriminant analysis and logistic regression analysis. As the accuracy of these failure prediction models is paramount, this paper sets out to compare these two statistical methodologies. Although not statistically significant, the conclusion is that logistic regression analysis produces slightly better predictive results than multivariate discriminate analysis.

Keywords:

• corporate failure
• multivariate discriminant analysis
• logistic regression analysis

Additional information

Notes on contributors

P W Court

P Court is with the Department of Business Administration at Rhodes University

S E Radloff

S Radloff is with the Department of Mathematical Statistics at Rhodes University