Good morning Alfonso,

I hope this message finds you well!

I have a few questions regarding my setup for MVPA analyses (resting state data for 3 subject groups, total of 87 subjects, ANOVA as a between-subjects contrast). I will be leaving the default 64 dimensionality reduction value. However:

1) I am not sure if I should be selecting the "Centering" option or not. Would you be so kind as to provide a brief description of what this does, and if it is recommended that I use it for my analyses (or not)?

2) Would you be so kind as to provide input as to which approach you think would be best?
Approach A: Extract 7 or 10 factors (around 1/10th of my sample size), and perform analyses on all factors.
Approach B: Extract 7 or 10 factors, but only analyse the first 3-4 (if there is a hierarchical nature to the factors as with ICA).
Approach C: Since most papers discuss 3-5 factors, simply extract 4 factors and investigate all of them.

3) Would you be so kind as to confirm that it is preferable to use an F test to analyse all the factors at once, instead of analyzing each factor individually (to mitigate for multiple comparisons)?

4) There seems to be no real consensus in the literature for the thresholding of significant clusters in the MVPA analyses. Some perform a cluster-level FDR, others use a cluster threshold of 50 voxels, others 100 voxels. In your expert opinion, is there a "best way" to proceed?


My apologies if anything wasn't clear.


Thank you very much, and have a wonderful day,
Maria