Why do we use t-test when one of its assumptions is violated? – ResearchGate 6.8.2015 | Pedro J Aphalo Source: Why do we use t-test when one of its assumptions is violated? Isn’t it better to use Mann-whitney U, when there is no equality in variances? – ResearchGate A crucial question, and as usual a spot-on answer by Jochen Wilhelm in ResearchGate’s questions and answers area.