Joint Inference of Baseline Hazard and Regression Coefficients in Cox-Like Models via Empirical Likelihood

• Activity Number: 22
• Type: Topic Contributed
• Date/Time: Sunday, August 4, 2013 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Mai Zhou*+ and Song Yang and Mi-Ok Kim
• Companies: Univ of Kentucky and NIH/NHLBI and Cincinnati Children's Hospital
• Keywords: Confidence Intervals ; Semiparametric ; Wilks LR test ; Prediction

Abstract

We use Cox proportional hazards model and Yang and Prentice (2005 Biometrika) model to illustrate how to use empirical likelihood ratio to make joint inference on the regression parameters and the baseline hazard function.

Cox's partial likelihood is free from the baseline, thus cannot be used for the inference that involve baseline. For that purpose an empirical likelihood seems more appropriate. Empirical likelihood for both models will be studied and compared to the partial likelihood in the case of Cox model.

Confidence intervals for the hazard ratio or hazard difference for two samples at given time points are studied. Some R software developed for carry out the said computation will be available.