Testing for Hardy-Weinberg proportions: have we lost the plot?

Abstract

Testing for Hardy-Weinberg proportions (HWP) is routine in almost all genetic studies of natural populations, but many researchers do not demonstrate a full understanding of the purposes of these tests or how to interpret the results. Common problems include a lack of understanding of statistical power and the difference between statistical significance and biological significance, how to interpret results of multiple tests, and how to distinguish between various factors that can cause statistically significant departures. In this perspective, which focuses on analysis of genetic data for nonmodel species, I 1) review factors that can cause departures from HWP at individual loci and linkage disequilibrium (LD) at pairs of loci; 2) discuss commonly used tests for HWP and LD, with an emphasis on multiple-testing issues; 3) show how to distinguish among possible causes of departures from HWP; and 4) outline some simple steps to follow when significant test results are found. Finally, I 5) identify some issues that merit particular attention as we move into an era in which analysis of genomics-scale datasets for nonmodel species is commonplace.