EXACT TESTS FOR RANDOM MATING IN AUTOTETRAPLOIDS

In statistical genetics, a common task is to evaluate the assumption of random mating, which is a baseline for population genetic models and can be used as a quality control procedure. In diploids, which are organisms with two copies of their genome, this isdone through tests for Hardy-Weinberg equilibrium (HWE). However, for autotetraploids, which are organisms with four copies of their genome, the assumption of random mating and HWE are different and there has been less work on evaluating random mating in these organisms. The standard approach for HWE tests in diploids is to use exact tests, which control for Type I error for finite sample sizes. However, there are currently no exact tests available for autotetraploids. In this study, we provide two approaches for exact tests for random mating in autotetraploids. The first approach conditions inference on sufficient statistics, while the second uses a split likelihood ratio. Although these approaches exactly control for Type I error, simulations show that they are too conservative for use. Therefore, we recommend using a standard likelihood ratio test or a standard chi-squared test, which we implement. We demonstrate all of these approaches using a dataset of autotetraploid white sturgeon.