0
$\begingroup$

I have estimated a Weibull regression model in BUGS/JAGS which gives me the log-hazard as a function of intercept (baseline hazard) and covariate effects.

The intercept is estimated as -9.826 and one of the covariates is estimated as 0.78. I can now compute the specific log-hazard for each individual in the dataset. However, I am interested in the expected survival time. Hence, I would like to predict each individual's survival time from their log-hazard.

I know that it is possible since, when estimating the same survival model in R, I can use predict(type="response") to get back to the survival times.

In my stats books and on the Internet, I can also find ways to express the survival probability but not the survival time. Is there a way?

$\endgroup$
0
$\begingroup$

I came up with the following idea: There is a link between the hazard estimates and the coefficients of an accelerated failure time parameterization of the Weibull regression model. Specifically, Beta_Hazard = - Beta_AFT / shape.

To get to the expected survival times, I would then first transform the regression coefficient to AFT estimates, and, in a second step, take the $\exp()$ of the linear predictor of the AFT model, which equals the survival time.

Can someone confirm that this works?

$\endgroup$
  • 1
    $\begingroup$ Yes, that will be correct to do. $\endgroup$ – Dimitris Rizopoulos Nov 16 '18 at 20:22

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.