Modification of wilks’ lambda test statistic in multivariate analysis of variance test

Abstract

This study modified Wilks’ Lambda Test Statistic in Multivariate Analysis of Variance Test. Simulation was done using R. The sample sizes utilized were 5, 10, 15, 20, 25, 50 and 100; Number of variables and Number of groups p = g = 2, 3 and 4 with equal, unequal samples sizes and variance co – variance matrix. The comparison were done at one level of α = 0.05 using power of the test and type I error rate. From the result obtained under type I error rate, it was indicated that when g = 2, p = 2, p = 3, p = 4, g = 3 p = 2, p = 3, p = 4, g = 4, p = 2, when g = 4, p = 3, when g = 4, p = 4 and compares with chi – square value. It was noticed that the proposed test statistic performs better at all sample sizes and across all variables when the sample sizes were large (i.e 50 and 100) but not perform better when the sample sizes are small. Under power of the test, it was obsrved that when g = 2 and p = 2, p = 3 and p = 4, g = 3 and p = 2, p = 3, p = 4, g = 4, p = 2, p = 3, p = 4 and compares all with chi – square value. It was noticed that the proposed test statistic does not performs better at all sample sizes and across all variables.