ON THE ESTIMATION OF THE PROPORTIONAL HAZARDS MODEL IN THE PRESENCE OF UNOBSERVED HETEROGENEITY

In this paper we discuss the consequences of omitted covariates for the estimates of the regression coefficients and the base-line hazard of the proportional hazards model. We find that the regression coefficients are underestimated and that the estimated base-line hazard is less increasing or more decreasing than the true base-line hazard.