Is an Alternative Method Always the Better One? A Comparison Between ANOVA, Welch's F, and the Kruskal-Wallis Test

• In Event: Poster Session 15
  In Poster Session: Accuracy and Meaning in Modeling

Abstract

Comparing three independent group means measured on a continuous scale, the question is typically whether to choose the traditional F test, Welch’s F test, or Kruskal-Wallis test especially when the statistical assumptions are violated. The current study compares the performance of traditional F test, Welch’s F test and Kruskal-Wallis test by examining the nominal error rates under varying degrees of non-normality, and heterogeneity of variance using a variety of sample sizes (20 to 200) and a balanced design. When homogeneity of variance is violated, Welch’s F test performs best under almost all such conditions. We found no benefit to perform Kruskal-Wallis test when one is concerned about failure to meet the assumptions of the traditional F test.

Authors

• Meng-Ting Lo, The Ohio State University - Columbus

• Yixi Wang, The Ohio State University - Columbus

• James Ohisei Uanhoro, The Ohio State University - Columbus