Compare Survival in Days using Mann-Whitney-U Test I already read across the CrossValidated StackExchange, but didn't find an answer on the following problem.
I have different groups of animals which have survival rates given in days since first contact, like [5, 10, still alive, 50, 104, still alive, ...] - how would I preprocess the data, especially the ones labelled with 'still alive', to do a Mann-Whitney-U test on the survival in days of two groups?
Would I first of all use Kaplan Meier method to censor the data, or can I transform the data in some kind of conditional probability, or would I just use the raw data skipping the ones which are still alive?
Thanks to anyone who can give me a hint on how to process this kind of data given the Mann Whitney U test.
 A: *

*Kaplan- Meier doesn't 'censor the data'! Your data are already censored -- this simply describes the fact that when you have survival times but some subjects are still living, their survival times are unknown but must be at least as long as you've observed them ... the survival times are censored at their observed time survived so far.
Kaplan Meier is one way of computing an estimated survival curve in the presence of such censoring.


*Be very careful about just coming up with some arbitrary approach to dealing with the data in the presence of censoring. You can't just throw censored data at whatever procedure and hope to get sense out. If you're going to do anything but the standard procedures for censored data (there are many such), you have to know how to deal with censoring properly.


*The Mann-Whitney(-Wilcoxon) is not suitable (as is) for censored data. It could be adapted -- e.g. to give upper and lower bounds on the p-value for a test of equality of survival time -- but it's better to simply apply procedures that already exist for comparing survival time. You might consider, for example, the logrank test, or a comparison via the Cox proportional hazards model, or a parametric survival model, for example.
You might find this document helpful as an explanation of Kaplan-Meier survival estimation and the logrank test.
