CMStatistics 2016: Start Registration
View Submission - CMStatistics
B0639
Title: Near optimum estimation of variance components in mixed effects models Authors:  Subir Ghosh - University of California (United States) [presenting]
Abstract: The variance component estimation for a mixed effects model is considered for analyzing the data obtained from an experiment or observational study. The optimum estimation method yielding the uniformly minimum variance quadratic unbiased estimator (UMIVQUE) is often impossible to carry out. A near optimum estimator (NOPE) is proposed in such situations under some distributional assumptions for the data. Examples of NOPE are presented in an example. A NOPE satisfies the unbiasedness property. The performance comparison of NOPE is made with the maximum likelihood estimator (MLE) and restricted maximum likelihood estimator (REMLE) by simulation for a designed experiment.