0

This feels like it should be a very straightforward stats analysis but I'm struggling to find a solution.

I have a data set made up of pre- and post-test continuous data values taken from multiple groups i.e.:

Group 1 (pre-test, post-test)

Group 2 (pre-test, post-test)

Group 3 (pre-test, post-test) etc..

I want to see in which groups there is a significant difference across the pre- and post-test measures. The most simple analysis would be a paired t-test, however I have many groups and am concerned that multiple paired t-tests will leave my results vulnerable to familywise errors.

An alternative might be to use a repeated measure GLM. I ran such an analysis on SPSS and sure enough came up with a between(group)*within(test-effect) interaction.

I'm now stuck as to how to find in which groups the within-subject comparison is significant. I can only think of splitting the groups up and running multiple comparisons in the estimated marginal means, with a Bonferroni correction. Only the Bonferroni correction does not work with just a single within-subject comparison.

Are there any kinds of family-wise error corrections I can do for the follow-up within-subject comparisons when there is only one-level of comparison?

Thanks!

user18071
  • 1
  • 1
  • 1
  • You can still do Bonferroni correction in SPSS -- but you have to do this manually. So e.g. Run four paired t-tests, and then compare the p-values to the new critical p-value set by Bonferroni correction: if you have four tests, then the new critical p-value would be .05/4=.0125. More complex post hoc comparison methods for within-subject comparisons exist, which I think are less conservative than this correction, but more complex to implement. See this question for some ideas http://stats.stackexchange.com/questions/14078/post-hoc-test-after-anova-with-repeated-measures-using-r?rq=1 – James Stanley Dec 24 '12 at 01:53
  • Thank you for your reply, I very much appreciate it! Happy new year! – user18071 Jan 01 '13 at 18:00

0 Answers0