0

How can I do a post-hoc test for survival analysis to reveal which groups differ significantly?

I have four groups (with 40 individuals of each) treatened by different chemicals. All of the individuals died at the end of the experiment, but the survivorship was significantly different in the four groups (I used "survdiff" command). However, I do not know which groups differed significantly from each other (for example I can use Tukey test for ANOVA as a post-hoc test, but I cannot find any test for this problem).

István
  • 1
  • 1
  • Differ significantly in what? Survival analysis produces several different numbers (e.g., $\hat{S}{t}$ and $\hat{h}{t}$ for many times "$t$", median survival time, mean survival time, etc.). Can you edit your question (click the "edit" link in the lower left) to indicate which of these and which comparisons (and what kind of contrast, difference, ratio, something else) you want to test? – Alexis Sep 11 '19 at 15:56
  • Also, welcome to CV, István Czeglédi! :) – Alexis Sep 11 '19 at 15:56
  • Thank you Alexis. I just want to know which chemicals are effective in killing parasites in this experiment. To this I compared the survival curves of the four groups with "survdiff" function (https://www.rdocumentation.org/packages/survival/versions/2.44-1.1/topics/survdiff) which indicated significant difference among them. I would like to know which groupes differed significantly. Should I test the median or mean time at death for this? – István Sep 12 '19 at 09:03
  • Difference in what? Effective in killing parasites when? At one point in time? Cumulatively? See my answer and comments on it to a similarly vague question here – Alexis Sep 12 '19 at 15:45
  • I just want to know whether there is a significant difference among groups in mean. And yes, cumulatively. – István Sep 16 '19 at 08:35

0 Answers0