0

I am comparing two groups (young vs. old adults) for differences in 5 brain regions (size) with independent t-tests. I also compare performance on three different aspects of a task with independent t-tests and then correlate brain sizes and behavioral performance . I adjusted my p-values (Bonferroni-method) for the correlations (for the five brain regions), but I am not sure, how exactly/if I have to adjust the p-values for each independent t-test between the age groups. I have a priori hypotheses about all comparisons. Any help would be highly appreciated!

Isabel
  • 3
  • 2
  • Welcome to Cross Validated, Isabel. Before asking _how_ to correct for multiple comparisons, you might consider whether you _should_ correct. See here for some discussion: https://stats.stackexchange.com/questions/3200/is-adjusting-p-values-in-a-multiple-regression-for-multiple-comparisons-a-good-i – Michael Lew Nov 21 '18 at 20:08

1 Answers1

2

If your study is preliminary, or if it is about characterisation rather than testing hypotheses then you probably should not `correct' for multiple comparisons. The raw P-values give you the strength of evidence against the individual null hypotheses and are should be presented alongside a more detailed presentation of the data.

If you decide that you should `correct' then you will have to decide whether to treat all of your comparisons as part of the family or whether your experiments consist of a couple of families. I would suppose that the brain size tests are one family and the behavioural tests are another.

Finally, you might consider a slightly less power-wasting form of multiplicity correction like one of the step-down procedures (search for Šidák and step-down).

Michael Lew
  • 10,995
  • 2
  • 29
  • 47