I need to perform survival analysis for 4 different groups in a study. When I use the Logrank test, the null hypothesis is rejected.

My question is, how can I tell now exactly which pair of groups causes the null hypothesis to be rejected? Is there a special post-hoc test, or should I repeat the logrank test manually for different pairs of groups? (or another solution, if possible)

Appreciate any kind of help!


Just do the two sample survival curve comparison test (logrank if that is what you like to use). But keep in mild the need to adjust for multiplicity just as you would with post hoc pairwise mean difference comparisons in an ANOVA.

  • $\begingroup$ Thank you! By that you mean, performing the Logrank test separately for each pair of samples (i.e, for 4 samples perform 6 different tests) while adjusting for multiple comparisons? $\endgroup$ – Shmulik Sep 15 '12 at 14:00
  • $\begingroup$ @Shmulik O mean do as many of the pairwise tests as you want. Just make sure you account for the multiplicity. $\endgroup$ – Michael Chernick Sep 15 '12 at 14:52
  • $\begingroup$ @MichaelChernick: Could you please be a little bit more specific about how to adjust for multiplicity ? What kind of correction do you have in the back of your head ? $\endgroup$ – ocram Sep 15 '12 at 15:42
  • $\begingroup$ Books have been written on this subject. There are many reasonable approaches including methods due to Tukey, Dunnett, Hochberg and several others (although some may be specific to comparing means). I like resampling type p-value adjustments such as bootstrap adjustment described by Westfall and Young in their book. $\endgroup$ – Michael Chernick Sep 15 '12 at 15:52

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.