standard error
On May 28, 2012, at 5:20 AM, Christopher Kelvin wrote:
Dear all, I want to determine the standard error or the mean squared error for the parameter estimate for beta and eta base on the real data. Any help on how to obtain these estimated errors. library(survival) d <- data.frame(ob=c(149971, 70808, 133518, 145658, 175701, 50960, 126606, 82329), state=1) s <- Surv(d$ob,d$state) sr <- survreg(s~1,dist="weibull") beta<-1/sr$scale p1=(beta) p1 eta<-exp(sr$coefficients[1]) b=(eta) b
The usual approach is to rely on the normality of the parameters on the "Weibull scale" and then back transform coef +/- 1.96*se(coef) You get these with summary(sr)
David.