Meta-Regression Models for Survival Data

Assuming that the individual level failure time follows an exponential distribution with mean 1/Xp the likelihood function of yp vf) is given by 

We consider both random effects and fixed effects models for in (2.2). The log-linear random effects model for assumes 

We see from (2.4) that under the exponential model, the likelihood function is solely based on the treatment-level meta-survival data D. We also see from (2.4) that under the exponential model, the likelihood function does not depend on the Ujfs. 

For simplicity, we take 0k = - 0. Unlike the random effects model, 

---