Power of pairwise multiple comparisons in the unequal variance case

Abstract

The Type I error rate and statistical power of six selected pairwise multiple comparison procedures were estimated by using simulated data in which population variances were heterogeneous. Based on a oneway Anova model, eleven different combinations of sample sizes and variance patterns were examined for a four-group design and another eleven patterns were examined for a six-group design. The results indicated that only the Games and Howell procedure occasionally failed to control the familywise Type I error rate under the nominal .05 level. In terms of statistical power, Shaffer (1979, 1986) enhancements to the Bonferroni approach provided the greatest average power per contrast and all pair power in the four group design. In the six-group design, no procedure consistently provided the greatest per-pair power or all pair power

Keywords:

• multiple comparisons
• Type I error rate
• any-pair power
• per-pair power
• all-pair power