I am trying to produce a series of plots showing the prevalence of a
condition, which is subject to censoring. In most cases the condition is
temporary and resolves with time. I would like to use the method of Pepe
et al Stat Med 1991; 413-421 - essentially the prevalence is the
Kaplan-Meier prob[having the condition at time t] - KM prob[recovery by
time t] (also divided by 1-KM[death by t], although death is not an
issue with this data).
I can easily produce the relevant actuarial data for either the
condition or recovery using survfit(eg survfit_cond$time ,
survfit_cond$surv, survfit_rec$time, survfit_rec$surv). I then have to
calculate (survfit_cond$surv-survfit_rec$surv) at each event time point.
Can anyone help me with an easy method to implement this? Or suggest an
easier method? I cant find a similar method after searching the
contributed packages (it doesn't appear to fit a recurrent events
problem). I have code for manual KM calculations, but the only method my
basic programming skills come up with seems tedious.