Survival analysis

Question

Suppose you were given survival data for subjects randomized to one of the two treatment groups. The variables included id, time (time to relapse in days), censor (1 if observing time to relapse, 0 if censored observation), and treatment (1 or 2). You estimated survivor function for each group using Kaplan-Meier method and compared the distributions by log-rank test. Now you want to re-run the analysis using month as the time scale; that is, use newtime=time/30.4. Will you get the same P value from the log-rank test after changing the time scale? Why?

My idea was the p-value will be the same because the time did not change just the time scale changed with do not have an effect on the survivor function. I am not sure if my idea is right.

Answer 1

But, does it not have an effect the survival function? What have you done to come to this conclusion. Have you works out a few numbers to see if your hunch is correct? I would think the sample size would also come in to play, but I don't see any information about that parameter.

I would also think about it in this way: if the time interval is longer, then you'll likely have more 'events', or deathes, per unit of time. Look closely at the equation for the Kaplan-Meier test (i.e. the log-rank test), If the

number of events goes up, what effect does that have on the test statistic. That is what ultimately gives you the p-value.

I can't do the work for you, but I've given you some other things to think about.