Inference for lethal gene estimation with application in plants

Recessive genes with deleterious effects have been of great interest to genetics researchers because these effects are critical to both plant and animal breeding programs for small closed populations. None of the previous methods in the genetics and forest genetics literature has allowed the development of statistical inference for the estimation of these effects. We propose an estimation method for the number of lethal equivalents, which is an overall measure of the genetic mortality, based on a hierarchical construction of the likelihood functions for mating experiments. Because of unobservable genetic variables in the likelihood functions, a variant method of moments, marginalized method of moments, is applied to derive these estimates. In particular, we are able to compute confidence intervals for the number of lethal equivalents. We illustrate our methods with two mating systems-selfing and full-sib crossing-using both simulated and real data.