Robust Procedures for Variance Component Problems Using the Jackknife

Abstract

It is well known (Scheff6 [1959]) that in components of variance problems the standard F tests are highly nonrobust against nonnormality. In many genetics problems, one is interested in obtaining estimates from what are essentially variance component models. The goal of this discussion is to propose consideration of the jackknife technique to obtain robust tests for these models. In addition, the technique can be adapted to obtain confidence intervals for parameters of interest. A Monte Carlo study is given, as well as an application to a quantitative genetics problem.