A Derivative-Free Approach for Estimating Variance Components in Animal Models by Restricted Maximum Likelihood1

Abstract

A derivative-free algorithm for use in restricted maximum likelihood variance component estimation for single traits and single records in animal or reduced animal models is presented. The algorithm does not require matrix inversion; instead, it uses a one-dimensional search involving the variant part of the log likelihood to find the maximum of this function. An approach to account for selection in base populations is also presented. Sample results are given.