MLE in the proportional odds model
Susan Murphy
Pennsylvania State University
Abstract
The proportional odds model is an important alternative to the
proportional hazards model for continuous time survival data. We can
estimate the regression coefficients by maximizing a profile
likelihood. This profile likelihood can be used much as a classical
parametric likelihood. That is, the maximum profile estimator of the
regression coefficient is asymptotically normal with efficient
variance. Consistent standard errors of the estimated regression
coefficients can be found by numerical differentiation of the profile
likelihood. As an alternate to the Wald test a profile likelihood
ratio test for the regression coefficient can be used..