Discriminant analysis with mixed non normal variables

Abstract

The mixed variable discriminant analysis procedure assumes that observations are distributed multivariate normal with different group means but same variance-covariance matrix. However, attention has not been given in discriminant analysis when the assumption of normality no longer holds. Therefore, we present a simple but new approach to mixed variable discriminant analysis when available observations (or its mixture) are not distributed multivariate normal. Specifically, a mixture of bernoulli and exponential and, poisson and bernoulli variates in discriminant analysis were presented in this work. Under a given condition, the suggested mixed non normal discriminant procedure demonstrated ability to allocate a mixture of non normal observations with minimal error.

Keywords:

• Non normal
• continuous
• discrete
• mixed variable
• allocation rule
• error rate

Additional information

Funding

This work was not supported by any grant.