Quadratic Information Criterion for Model Selection in Nonlinear Mixed Effects Models

• Activity Number: 568
• Type: Contributed
• Date/Time: Wednesday, August 1, 2012 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistical Learning and Data Mining
• Author(s): Rositsa B Dimova*+ and Marianthi Markatou
• Companies: Weill Cornell Medical College and IBM T. J. Watson Research Center
• Keywords: Model Selection ; Quadratic Distances ; QICh ; Mixed Effects Models

Abstract

Common approaches for assessment of goodness of fit for mixed effects models are based on the likelihood function, the most popular of which is the AIC. The question we investigate here is whether we can use the quadratic distances framework for model assessment in mixed effects models. Quadratic distances between probability distributions as measures of goodness of fit were discussed in Lindsay et al. (2008) and Lindsay, Markatou and Ray (2011). In our previous work (Dimova and Markatou (2012)), we derived a criterion for model selection, called the quadratic information criterion (QICh) for nonlinear regression models and studied its asymptotic behavior and performance. In this work, we derive the QICh for the nonlinear mixed effects model with fixed matrix of the random effects. We investigate different tuning parameters of the distance kernel and study the behavior of the proposed criterion through simulation experiments. Our simulation experiments illustrate that for appropriate kernel and tuning parameter, the QICh criterion outperforms the standards for model selection AIC and AICc.