Dear all,
Happy New year!
I have used the 'crr' function to fit the 'proportional
subdistribution
hazards' regression model described in Fine and Gray (1999).
dat1 is a three column dataset where:
- ccr is the time to event variable
- Crcens is an indicator variable equal to 0 if the event was achieved, 1
if the event wasn't acheived due to death or 2 if the event wasn't
achieved
due to disease progression
- pre is an indicator variable (and the covariate of interest)
I want to investigate if pre has a significant impact on time to event for
patients who died and for those who suffered disease progression (as well
as it's impact on the overall time to event).
The code I have used is as follows:
fitd <- crr(dat1$ccr,dat1$Crcens,dat1$pre,failcode=1,cencode=0)
fitp <- crr(dat1$ccr,dat1$Crcens,dat1$pre,failcode=2,cencode=0)
In these cases I get p-values of 0 and 0.66 respectively.
What I would now like to do, is to plot two cumulative incidence curves -
one for the 'pre' variable status for patients who didn't acheive
the event
due to death and one for those who didn't achieve it due to progression.
How can I do this? I can only see things involving plot.predict.crr which
doesn't seem to be what I need?
Many thanks,
Laura
Usung Windows 7 and R 2.14.1
[[alternative HTML version deleted]]
