Thank you to the various people who have made suggestions. In particular, reading the documentation of the addreg package has prompted me to try to put the question differently. I would be grateful for any comments on the following.
As I described before, I have a snapshot of a population taken at a certain time. I am interested in an age-related disease, which progresses healthy->A->B. (There is no recovery.) For each individual, I know their age (in years) and the stage of the disease. There are roughly 800 cases, with ages spanning 40 years. Suppose I don't distinguish between stages A and B, and all I am interested in is whether someone has the disease or not. For each individual, I therefore have a censored observation of a "lifetime" random variable: if the individual is age t and is diseased, lifetime is in (0,t]. if the individual is age t and is healthy, lifetime is in (t,inf) I would like to plot a survival function for this "lifetime" random variable. According to the documentation (for R1.7.0), the Surv function does not let me enter left-censored intervals for non-parametric plots. Are there ways around this? I could simply estimate Prob(lifetime>t) = fraction of cases of age t who are healthy and take this as my survival curve, but it produces a noisy plot (in particular, the curve is not monotone). Is there a good way to get a better estimate of the survival function? Once I have a good way to estimate survival functions for this sort of data, I could estimate the distribution of T1 (the time to reach stage A or B) and of T2 (the time to reach stage B), and thereby estimate the distribution of T2-T1 (the time to progress from stage A to stage B) by some sort of convolution, assuming independence. Damon. ______________________________________________ [EMAIL PROTECTED] mailing list https://www.stat.math.ethz.ch/mailman/listinfo/r-help
