Dear Experts and Colleagues,
In my recent research, I encountered a statistical issue: in an initial exploratory MEG analysis, I performed analyses for the entire frequency bands of delta, theta, alpha, beta, gamma1, and gamma2 between groups. However, only the theta band showed strong positive results, while the other frequency bands either showed no effects or yielded poor outcomes. Therefore, after applying FDR correction to control for both the frequency and signal dimensions, my results turned out negative. In this situation, would it be acceptable not to correct for the frequency dimension, but rather treat each frequency band as an independent hypothesis and apply corrections solely for the signal dimension within each frequency band?
Thanks for your help.
If your research question includes all frequency bands, you should take them into account; otherwise, this would be a form of p-hacking. Of course, it's an initial exploratory analysis, so there is something to say about that, but if this data is to be published, then I wouldn't favor a post-hoc selection of frequency bands and try to get in more subjects to improve the statistical power.
Dear SBeumer,
Thanks for your reply. Anyway, there are really rare answers online for this question. I would consider that more cautiously in my further research.
Nice day 