# Is there a repeated measures aware version of the logrank test?

I am looking for a repeated measures version of the Logrank test. If I am correct, I am looking for an equivalent of the Friedman test for survival data.

Any suggestions on where to look? (and R code will always be welcomed :) )

Thanks.

• More information on the nature of the repetitions would be helpful. Perhaps clustered survival data methods including frailty models apply (and perhaps not). – Aniko Apr 27 '11 at 19:16
• There are n subjects, each one has k tasks timed. I wish to compare the timing of the tasks between each other (are there timing different or not). But they are dependent through the subject. (Thanks) – Tal Galili Apr 27 '11 at 20:52

You can probably use a so-called marginal survival model. It would be more like Cox-regression than a log-rank test, i.e. proportional hazards for the effect of task would be assumed. It is implemented in the survival package:
mod <- coxph(Surv(time, censor) ~ task + cluster(id))

Here task would be a factor representing the task, and id would identify the subject. There is a book that goes into lots of detail on multivariate survival data: T.Therneau, P. Grambsch. Modeling Survival Data: Extending the Cox Model. Springer, 2000.