Investigating the Impact of Nonnormality, Effect Size, and Sample Size on Two-Group Comparison Procedures: An Empirical Study

• In Event: Alternative Methods for Data Analysis

Abstract

For two-group comparisons in which the assumption of normality is not tenable, previous researchers have specifically emphasized that the Wilcoxon rank-sum test provides greater power compared to the independent samples t-tests. Most empirical studies have examined robustness and power when the two populations show the same pattern of non-normality. Given this, the current simulation study examines Type I error control and statistical power of the independent samples t-tests, the Wilcoxon rank-sum test, and nonparametric bootstrap under group-specific non-normal distributional conditions, various sample sizes, and effect sizes found in educational research. Recommendations will be provided and implications for practice are discussed.

Authors

• Yoonjeong Kang, University of Maryland - College Park

• Jeffrey R. Harring, University of Maryland