Determine Robust Procedure for Testing Variance Equality Using Type l Error Rate and Power

Comparisons were made between seven of the many procedures used to determine variance homogeneity. The seven tests that have been chosen are the Bartlett's test, the Levene's test (mean), the Levene's test (median), the Levene's test (trimmed mean), the O'brien test, the Cochran test, and the Fligner's test. Data were simulated in order to compare the seven procedures using different distributions (Normal, Beta, and Uniform), sample sizes (5, 10, 50, and 100), equal samples (n_1=n_2=...=n_k), and five (5) levels (i.e., k = 5). The power of the test and type l error rate were used to compare the selected procedures at a significant level of 0.05. The findings demonstrated that the Fligner technique is superior to all other procedures when the dataset is normally distributed, whereas the Bartlett procedure is superior regardless of sample size when the dataset is not normally distributed.