An alternatif test to the two independent samples t test

TL;DR

This paper introduces a new statistical test based on cross variance as an alternative to the traditional two-sample t-test for comparing means of independent normally distributed samples, especially when variances are homogeneous.

Contribution

The paper proposes a novel cross variance-based test and demonstrates its comparable power and error rates to the t-test through simulation studies.

Findings

Cross variance test has similar power to the t-test.

Error rate of the cross variance test matches the t-test.

Suitable as an alternative for mean comparison in normal samples.

Abstract

In this paper we proposed the alternative test to the two independent and normally distributed samples t test based on the cross variance concept. We present the simulation results of the power and the error rate of the special case of the cross variance which is when the variances of the two samples are homogeneous and the t tests. The simulation results show that the special case of the cross variance test has the power and the error type I rate equal to the $t$ test. This result suggests that the proposed test could be used as an alternative to detect whether there is difference between means of the two independent normally distributed samples. We give some example comparative case studies of the special case of the cross variance and the t tests.