We have two groups of participants -- experimental and control. The experimental group was provided an intervention to help them deal with social anxiety. The control group was given a list of on-campus resources they can contact when stressed. We measured the experimental and control group participants on a group of measures at pre- and post-intervention -- Distress tolerance scale, Intolerance of Uncertainty scale, Confidence scale (which has 3 sub-scales), and Intention to persist in science scale. Based on multiple linear regressions, we observed that the experimental group participants had increased confidence (on all 3 sub-scales) at post-intervention, and the effect of the group is significant (p
1) Multiple comparison correction must be applied for *all* comparisons, hence here for your 6 p-values if you have made 6 tests (and are interested in the results of the 6 — interested meaning that you decided to perform the comparison, instead of an automatic result of a software. But of course, if you look at the p-value to decide if your comparison is finally interesting of not, then you *must* include this p-value in the correction.
2) Note that FDR controls the false positive rate, but does not control the risk that none of your results are false positives. If you want to control this risk, you should use FWER approaches instead.
