[R] frailty

hsr hsanz at imim.es
Tue Apr 5 10:49:42 CEST 2011


Hi R-users

I spend a lot of time searching on the web but I didn’t found a clear
answer.
I have some doubts with 'frailty' function of 'survival' package.

The following model with the function R ‘coxph’ was fitted: 

modx <- coxph(Surv(to_stroke, stroke) ~ age + sbp + dbp + sex +
frailty(center,distribution = "gamma", method='aic'), data=datax)

Then I get survival (eg to 10 years) for the mean of the covariates:

survfit1 <- survfit(modx)
timesele<- 3652.25
tab <- as.data.frame(cbind(survfit1$time, survfit1$surv))
names(tab) <- c("time", "surv")
meansurv <- tab[tab$time==max(tab$time[tab$time<timesele]), ]$surv

‘survfit.coxph’ allows the entry of 'newdata' with the characteristics of
individuals that I want. This function doesn’t allow entering a value of
'center' (frailty term). However, when it performs the linear predictor
(modx$linear.predictors) it takes into account the frailty calculated for
each ‘center’ and applies to the value of the individual. 

My thought is: By extracting the average survival is not necessary to
consider the frailty term performed for each ‘center’ because the expected
value of the frailty terms is 0 (some centers will have more frailty and
other ones less). However, as each individual belongs to one ‘center’ it is
needed to take into account the frailty of that ‘center’ to perform the
linear predictor.

My doubt is: How the ‘survfit.coxph’ calculates the survival function with a
frailty term? 

Am I right with my thought ?
Can anybody help me with my doubt?

Thanks in adavance.





--
View this message in context: http://r.789695.n4.nabble.com/frailty-tp3427452p3427452.html
Sent from the R help mailing list archive at Nabble.com.



More information about the R-help mailing list