Dear SPM community,
I have a question regarding the corrections for multiple comparisons of post hoc tests, where I have a significant 3 x 2 interaction for multiple ROIs. We extracted the Cerebral Blood Flow (CBF, using the ASLtbx toolbox) of 10 ROIS in different conditions (e.g., CBF before and after consumption of 2 different drinks for two visits). We then separately calculated a linear mixed effect model in R, for each ROI, with several factors and covariates as fixed effects and the subject as random intercept. For significant interactions (e.g., Body Mass Index with 3 levels (lean, overweight, obese) and sex with two levels (female/male) results in a 3 x 2 interaction). We then used the emmeans function in R for post hoc tests and Bonferroni-Holm as correction procedure. Let’s say we found a significant interaction effect in three ROIs. Now to my question:
Do I now still have to correct the already corrected p-value for the number of ROIs?
I would like to avoid including the ROI as an additional factor, since the file is very bloated in the long format for statistical analyses. What is the correct procedure for such a multiple comparison problem with multiple ROIs and post hoc test for significant interactions?
Best
Ralf
|