search for: timeto

...s from Keiding and Klein artricle published in the Statistics in Medicine 1996) . While the simple weibull model runs perfectly ( I get -2*loglikelihood 1093,905), the weibull frailty model does not run well. The code I write is : likelihood.weibul <- function(p) { cumhaz <- exp(x1*p[4])*(timeto^p[2])/p[1] cumhaz <- sum(cumhaz) lnhaz <- (x1*p[4]+log((p[2]*timeto^(p[2]-1))/p[1]))*stat lik <- p[3]*log(p[3])-log(gamma(p[3]))+sum(lnhaz)+ log(gamma(di+p[3]))-(p[3]+di)*log(p[3]+cumhaz) -2*lik } di <-sum(stat) initial<-c(500000,2.8,.5,1.8) # initial 1/exp(constant) t<-nlm(likeli...