Abstract:
|
Testing for Hardy-Weinberg Equilibrium (HWE) is routinely conducted as an initial screening for genotyping errors in the analysis of single-nucleotide polymorphism. Much less effort, however, is spent in the estimation of the inbreeding coefficient, which can be considered as a measure of the deviation from HWE. For biallelic single locus, we use Monte Carlo simulations to compare the following four estimators of the inbreeding coefficients in terms of bias, mean square error, average interval length, coverage probability, as well as the probability of being outside of the parameter space. They are direct sample estimator, maximum empirical likelihood estimator, parametric Bayesian estimator and Bayesian pseudo-empirical likelihood estimator. The simulations show that both frequentist methods are similar and may produce estimators outside of the parameter space when the allele frequency is small. On the other hand, the Bayesian approach always produces estimator within the parameter space through an appropriate prior distribution. Furthermore, the Bayesian estimators have larger coverage probability and smaller MSE in the exchange of larger bias and longer interval length.
|