Prediction of colon cancer expense via adaptive penalized mixed model selection

Abstract

To predict the expense of colon cancer, we model it by linear mixed models, which consist of both fixed effects and random effects. In the mixed modeling setting, to identify the most appropriate model for real life data, we develop a two-stage procedure based upon the adaptive penalty term to separately select the suitable random effects and fixed effects. In the first stage, the random effects are chosen by means of the penalized restricted profilel log-likelihood; in the second stage, the fixed effects are selected through the penalized profile log-likelihood after the random effects are determined. In each stage, the Newton-Raphson algorithm is employed to implement parameter estimation. A simulation study is carried out for demonstrating the effectiveness of the proposed procedure. The proper model for the expenditure on colon cancer is ultimately selected with an application of the proposed method on a colon cancer data set.