Covariance Selection in the Linear Mixed Effect Mode

[edit]

Jonathan P. Williams, Ying Lu ;
Proceedings of the 1st International Workshop on Feature Extraction: Modern Questions and Challenges at NIPS 2015, PMLR 44:277-291, 2015.

Abstract

This paper improves and extends the two-step penalized iterative estimation procedure for the linear mixed effect model (LMM) by explicitly penalizing the off-diagonal components of the covariance matrix of random effects. To explicitly penalize the off-diagonal terms in the covariance matrix of random effects, glasso is incorporated in the penalized LMM approach. The paper also provides theoretical justification and a computational algorithm for the provided approach. Empirical analysis using random simulated data shows that explicitly penalizing the off-diagonal covariance components can greatly improve the model selection procedure.

Related Material