ABSTRACT
Multivariate distributions and related techniques are becoming increasingly popular with the progress of computing facilities. The continuous multivariate distributions include, among others, the multivariate normal, the beta distribution, and the multivariate t distribution. A justification of using different types of multivariate distributions is that, in many situations given a set of data, the assumption that they arise from a multivariate normal population does not always hold. It is mainly due to the fact that among all known multivariate distributions, it is the simplest one to handle and many well-known optimum statistical techniques of univariate normal have been successfully extended to the multivariate normal case. An important characteristic property of the multivariate normal distribution is that any linear combination of its components is normally distributed. In analyzing multivariate data, the role of multivariate normal distribution is utmost important. In actual practice, however, the assumption of multinormality does not always hold and the verification of this assumption is cumbersome, if not, impossible.
