Hi there,
I am working with an EEG dataset that looked at alpha, beta and theta frequency range in 4 brain regions: frontal, parietal, temporal and occipital. Participants performed two tasks: task A and task B. I am interested in the task effect, hemisphere effect as well as interaction effect in each brain region.
As such, I conducted repeated-measures ANOVA for each brain region in each frequency band, i.e. 4 brain regions x 3 frequency bands = 12 ANOVAs in total for this analysis.
As such, would I need to apply a correction to the resulting p-value, or p-value adjustment is not necessary as the ANOVAs are done in different brain regions?
Thanks in advance!
P-values are garbage. With that said, most researchers would probably advise that you do make a significance adjustment. As it is, you have a 45.9% chance of committing a Type I error. It doesn't matter that the tests are for different brain regions or frequency ranges.
My personal recommendation is to report ALL F statistics and confidence intervals for each regardless of statistical significance since these statistics are more informative than p-values when deciding which effects are important.
Blaine Tomkins is correct, you will have a huge chance to get a Type I error. The other option is to simulate your data and build a distribution of surrogate numbers. Then you can compare your actual numbers with the distribution. Moreover, if you are going to publish this, your simulated approach will improve the innovation of your paper.
- Badges
- Science topic