Online Program Home
My Program

Abstract Details

Activity Number: 495
Type: Contributed
Date/Time: Wednesday, August 3, 2016 : 8:30 AM to 10:20 AM
Sponsor: Biometrics Section
Abstract #321305 View Presentation
Title: A Computationally Efficient Algorithm for Random Effects Selection in Linear Mixed Models
Author(s): Mihye Ahn* and Helen Zhang and Wenbin Lu
Companies: University of Nevada, Reno and University of Arizona and North Carolina State University
Keywords: Lasso ; Linear approximation ; Nonnegative garrote ; Variance component selection ; Linear mixed model

The random effects selection has been received little attention in the literature. In linear mixed models, several methods for random effects selection have been proposed. However due to computationally intensive tasks, it is limited to apply the existing methods in practice. In this paper, we propose two approximate methods of the moment-based method for random effects selection. The exact moment-based method has two challenging computation issues: nonlinear semidefinite programming and nonlinear programming with a linear inequality constraint. In particular, the most time-consuming step is the second computation to produce sparse solutions of the variance-covariance matrix of random effect factors. Since the objective function has up to fourth order terms and it makes the computation tedious, we suggest using a linear approximation to the penalized variance-covariance matrix. It reduces the objective function up to second order, and the quadratic programming can be easily implemented in some statistical software. By simulation studies, we show that the approximate methods also perform well and often outperform the exact method.

Authors who are presenting talks have a * after their name.

Back to the full JSM 2016 program

Copyright © American Statistical Association