Investigation of the robustness of the student's t-test under the violation of the assumption of equality of variances

Abstract

The robustness of the Student's t -test is investigated under the violation of the assumption of equality of variances. With the aid of computer simulation, Type I and Type II error rates and the resulting statistical inference are studied and the effects of unequal variances on rejection rates and the power of the test are determined. Limits are determined on the degree of violation of the equality of variances that still leads to a satisfactory result when Student's distribution is used.