Abstract:
Hotelling's T2 statistic is one of the multivariate tests that are based on
certain fundamental assumptions, one of which is the multivariate normality. The test is
usually applied to investigate the equality of means associated with two normal
multivariate populations. The application of the test when the normality assumption is
violated may affect the power of the test. In this paper, a Monte Carlo simulation is
conducted to compare the exact powers of T2 based on multivariate normal and four
multivariate non-normal populations. The results show insignificant differences in the
average power performance between the normal and non-normal populations for both
small and large sample sizes. It was also observed that the results from multivariate
normal distribution do not show any unique performance from non-normal distributions
when samples sizes are large.
