By Thomas J. Faulkenberry (Tarleton State University); Ruth Horry (Swansea University)
Abstract
Though touted as one of the primary assumptions of the pooled independent samples t-test, the role of equal variances is often left unclear and under-investigated in classroom settings. When asked about the role of this assumption, teachers may be forced to provide only a superficial answer. To satisfy our own curiosity about the equal variances assumption, we developed an interactive web application (a Shiny app) and an accompanying activity to help students explore the impact of unequal variances on the t-test. The app allows users to see the distribution of p-values obtained from various combinations of effect size, population variances (equal, not equal), and sample sizes. We leverage this visualization in a corresponding classroom activity, where we specifically focus on how unequal variances can greatly inflate the Type I error rate, allowing students to immediately see for themselves the need for the equal variances assumption in the pooled independent samples t-test. In our session, we will demonstrate this app and share the classroom activity, which we have previously used with great success with first year psychology students.