Abstract:
The objective of this research was to study and to compare the efficiencies of nonparametric statistical test for two independent difference medians. The 5 statistics were Wilcoxon-Mann-Whitney Test (WMW), Fligner-Policello Test (FP), OGorman Adaptive Test (OG), Brunner-Munzel Test (BM) and Modified Intrinsically Ties Adjusted Mann-Whitney U Test (MWU). The two populations included logistic distribution, laplace distribution, gamma distribution and lognormal distribution. The ratios of variance were 1, 3, 5 and 7. Flypothesis testing were set at significant level of 0.05 and 0.01.
The results revealed that when the population has logistic distribution, variances were equal and unequal, sample sizes were equal and unequal, FP has the highest power of a test and variances were unequal, sample sizes were unequal, BM has the highest power of a test and sample sizes were equal, WMW has the highest power of a test. When the population has Laplace distribution, variances were equal and unequal, sample sizes were equal and unequal, FP has the highest power of a test and variances were unequal, sample sizes were equal and unequal, WMW has the highest power of a test. When the population has Gamma distribution, variances were equal, sample sizes were equal and unequal, BM has the highest power of a test and variances were unequal, sample sizes were unequal, OG has the highest power of a test and sample sizes were equal, BM has the highest power of a test. When the population has lognormal distribution, variances were equal, sample sizes were equal and unequal, OG has the highest power of a test and variances were unequal, sample sizes were equal and unequal, FP has the highest power of a test. When the sample sizes increase, most statistics tend to increase the power of the test rise.