Year: 2025
Type: article
Abstract: In psychological research, the independent samples t-test (Student’s t-test) is widely used to compare means of two independent groups, but its assumptions of normality and equal variances are often violated, compromising Type I error control. Welch’s t-test, which relaxes the equal variances assumption, provides a robust alternative yet is underutilized despite its advantages. Synthesizing evidence from simulation studies and empirical surveys, this article demonstrates Welch’s t-test’s superior performance in maintaining nominal Type I error rates under unequal variances and sample sizes, with negligible power loss compared to Student’s t-test. It also outperforms the Mann-Whitney U test in heterosce-dastic conditions. Despite its ease of implementation in statistical software, surveys indicate infrequent use and inadequate variance checks in published research. This paper advocates for Welch’s t-test as the default method for two-group comparisons in psychology, supported by practical examples and recommendations for educational reform to enhance statistical rigor. This paper was written with the assistance of Grok 3 an artificial intelligence assistant.
Source: Sociology & Social Policy
Author: Robert W. McGee
Institution: Fayetteville State University
Language: English
Subfield: Statistics and Probability
Field: Mathematics
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7128501096