Hi,

I've been wondering about this question for a while now. I've read up on the infamous multiple comparison problem and its appearance in analyses of variance. I think I understand the core principle: one runs a certain risk of making a type I error, and this risk increases when doing multiple group comparisons using the same sample. Bonferroni corrections (or other methods) are used to compensate for the increase in probability. Probability goes up, alpha must come down.

I am currently dealing with the above question in relation to an assignment: Does one run the same risk when multiple hypotheses are formulated in preparation of performing data analyses? And are Bonferroni corrections necessary to compensate for an increased risk of a type I error? And what if this is not a comparison of groups, but a test of R-square change?

For my assignment I have formulated one central research question (and corresponding central hypothesis). For testing, five hypotheses are formulated. Four hypotheses use the same predictor (x1) variable to predict four different outcome variables (y1, y2, y3, y4). One hypothesis uses a different predictor variable (x2) and another different outcome variable (y5). All tests are performed on the same dataset (N ~ 400).

Logically, I would think that the same principle would apply and Bonferroni corrections are necessary, however I never see it talked about in this context. So I am at a loss. Any input would be greatly appreciated, thanks!