Dear all, As part of the output for a t-test, SPSS performs a test of equal variances and then gives two results for the t-test depending on whether this first test fails or not. My question is whether you should base the equality assumption on this test or if you should decide beforehand whether the two populations would be expected to have equal variances. Consider the situation where one of the groups is very small - the estimate of the variance in this small group may be rather poor with the result the equal-variance test fails (especially if the combined sample size of 10,000 is swamping the test with power). Any thoughts? cheers Jon -------------------------------------------------- Jon Heron, PhD Research Statistician Avon Longitudinal Study of Parents and Children [log in to unmask]