I measured one dependent variable between three groups of data (control, group1, group2).The result of t-test indicates significant difference between control and group 1 (p vale =.04), control and group 2(p value .01), but not significant between group 2 and group 3.Do I still need to use the post hoc test? If so, which test is the most appropriate?

  • $\begingroup$ Kamran, I suspect you're at the wrong site for this question and are unlikely to get a good answer. We don't currently have many mathematicians here who are interested in statistics, but even if we did this question looks more like a practical question about scientific conventions rather than a theoretical question of interest to mathematicians. A similar question mathoverflow.net/questions/22532 did not get any answers. You and Pat could discuss whether there's a community for a statistics SE site, but I do not think MO is the place for that since we're mathematicians not scientists. $\endgroup$ – Noah Snyder Apr 26 '10 at 21:51
  • $\begingroup$ There is a proposal for a statistics SE site here: meta.stackexchange.com/questions/5547 $\endgroup$ – Rob J Hyndman Apr 29 '10 at 1:05

If you've got more than two groups, don't use t-tests (due to the multiple comparisons trap). Use a one-way ANOVA instead (based on an F-test). Then if you want a post-hoc test after that, use Tukey's HSD.

| cite | improve this answer | |

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.