I like to model censored events by subject cohort, comparing measured and predicted arrival times to see if there are significant differences by cohort.
It seems like a COX model might be appropriate, however I am not clear how the baseline curve is being generated. It seems the baseline is by default the factor 0 and not the non-stratified.
Ideally, I would like a baseline that is generated from all data points, irrespective of cohort, and generate a report of cohort specific predicted and measured expected arrival times, event counts by a certain time for each cohort and in comparison with baseline, along with the respective confidence intervals.
Any helpful hint or related reference would be highly appreciated.
Thanks
Sample started code, not that cohort A is more recent (shorter time history) compared to cohort B and C. Ideally I like to be able to use the model to extrapolate all cohorts to the overall longest available time.
library(survival)
data <- data.frame(
cohort = rep(c("A", "B", "C"), times = c(3, 6, 9)),
time = c(1, 2, 3, 1, 2, 3, 4, 5, 6, 1, 2, 3, 4, 5, 6, 7, 8, 9),
status = c(1, 0, 1, 1, 0, 0, 1, 0, 1, 1, 1, 0, 1, 0, 1, 1, 0, 1) # 1 = event, 0 = censored
)
cox_model <- coxph(Surv(time, status) ~ cohort, data = data)
summary(cox_model)