I am analyzing sage-grouse survival using cox proportional hazard models and need to test my global model for overdispersion by calculating a c-hat. Most people use the c_hat{AICcmodavg} function; however, this function doesn't seem to recognize the object produced by coxph models (see below). Is there any other similar functions that can produce a c-hat value for a cox's model? Thanks, Kyle

KE_global <- coxph(Surv(Start, End, Event) ~ Age+Res+Sex+Year1+DDM, data = sagr)

c_hat(KE_global, method = "pearson")

Error in c_hat.default(KE_global, method = "pearson") : Function not yet defined for this object class


I am not sure, and I havn't seen overdispersion done on cox hazard models before. Perhaps it isn't very surprising that this error occurs. In the official package documentation it states that:

Functions to compute an estimate of c-hat for binomial or Poisson GLM’s and GLMM’s using different estimators of overdispersion.

Checking for over- or under-dispersion is usually performed when the obsereved variance is greater than expected. In overdispersion :

A common task in applied statistics is choosing a parametric model to fit a given set of empirical observations. This necessitates an assessment of the fit of the chosen model.

The Cox proportional hazard model is, however, semi-parametric. More importantly, it does not assume any underlying distribution in order to fit a model. See Cox Proportional-Hazards Regression for Survival Data / John Fox:

This model is semi-parametric because while the baseline hazard can take any form, the covariates enter the model linearly.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.