Assessing the Effectiveness of Non-Parametric Methods for Comparing Patient Groups in Psychological Studies

Jeronimo Usuga-Gomez, Universidad EAFIT

Co-authors: Olga Usuga-Manco, Universidad de Antioquia; Freddy Hernandez-Barajas, Universidad Nacional de Colombia

Abstract: This study evaluates the performance of three non-parametric tests (Wilcoxon, Studentized Wilcoxon, and Brunner-Munzel) used in psychological research for comparing two independent groups. Using a comprehensive simulation study, we found that the Brunner-Munzel test exhibits superior statistical power for small sample sizes. To illustrate its practical utility, we apply all three tests to a real-world psychological dataset investigating whether optalidon (a drug) induces anxiety in patients during regular use. Our findings reveal no statistically significant difference in anxiety levels between optalidon and placebo groups across all tests.