Robust Estimation for Finite Mixture Regression Models with Random Effects via the Minimum Hellinger Distance Approach

• Activity Number: 401
• Type: Contributed
• Date/Time: Wednesday, August 6, 2008 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Liming Xiang*+
• Companies: Nanyang Technological University, Singapore
• Keywords: Kernel estimator ; Minimum Hellinger distance ; Finite mixture ; Residual maximum likelihood estimation ; Robustness

Abstract

The finite mixture regression with random effects is effective in modeling simultaneously the heterogeneity for clustered data arising from several latent subpopulations and the dependency among the data. In the literature, the residual maximum likelihood estimation (REML) has been studied for the model. However, the REML estimators tend to be unstable in the presence of outliers or extreme contamination. This paper focuses on developing robust estimators using the minimum Hellinger distance (MHD) approach. Under certain conditions, the MHD estimators are shown to be consistent and asymptotically normal. Monte Carlo simulations show that the MHD estimators perform satisfactorily for data without outlying observation(s), and outperform the REML estimators when data are contaminated. Application to a data set of neonatal length of stay (LOS) is presented to illustrate the method.