My dataset consists of more than 90% of Censored observations and 10 % of events in the observed duration. The problem I encountered while applying
was that the survival curve obtained never reached the 50% probability and it reached a 90% probability after a short span of time and went on like that for the rest of the duration.
But if I put
The survival curve looked normal with the curve reaching near 0 probability over time.
Now the problem arose when I was trying to find out individual probabilities on certain time instances and there I found that for a few observations, the probability increased to an extent and then reduced. But this is a clear anomaly?
My question is what is the interpretation statistically of
Also does this create a survival object same as if Event argument had been left as it is? In R Manual I could not find enough reference but I could find references where
Surv(Time,Event==1) is taken to analyse the data where Event has occurred
Let us give you some context:
library() # see the list of available packages library(survival) # load it. You can also # click the pull-down manual for packages and load it. library(help=survival) # see the list of available functions and data sets. data(aml) # load the data set aml aml # see the data #stime <- c(2.2, 3, 8.4, 7.5,9.6,8.5,0.2,1.7,1,0.1) #status <- c(1,0,0,0,0,0,0,0,0,0) Surv(aml$time, aml$status==0) coxfit1<-Coxph(Surv(time, status==0)~x, data=aml) survfit(coxfit1)#Could be written like this `survfit(coxfit1, newdata=data.frame(x=1))` as well to obtain the survival function of a particular subject with specific covariate values (x = 1) basehaz(coxfit1)
Now I want to obtain the individual probabilities at specific time instances t_1 and t_2