Exact significance levels of goodness-of-fit tests for the Hardy-Weinberg equilibrium.

The aim of this work is to examine the validity of several chi2 criteria, when used to test the fit of genotypic frequencies to the Hardy-Weinberg law. The exact type I error probabilities were compared to the nominal significance levels for two genetic models. The Pearson's chi2 test usually provided the best approximations. This test was found to be remarkably robust, when in situations involving very small expected frequencies.