Abstract
Maximum Likelihood (ML) and Restricted Maximum Likelihood (REML) estimators of variance components are widely used in animal breeding. Hoeschele (1989) has stated that the likelihood is always unimodal for REML, but this statement is not true in all cases. The surface of the profile likelihood functions for ML and REML are discussed and compared for a general two variance components model and for the one-way ANOVA model. For some cases upper bounds for the probability of multimodality are given.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have