 |
 |
Hello, We have a question regarding eigenvariate extraction for PPI analysis using SPM12 (as well as thresholding issues). What would be considered the most methodologically sound method of extracting eigenvariates? At the moment we are using a fairly liberal threshold to obtain the statistical map (p=.3-p=.5 at the individual subject level) and then extracting the eigenvariate from the nearest coordinates of peak activation to our ROI’s coordinates. However, it has come to our attention that masking out the non-relevant activation followed by then extracting the eigenvariate from the nearest peak coordinates might be a better approach. However, what is unclear to us is, if to obtain the SPM results, we already mask the ROI (e.g., 6 mm sphere in the dlPFC) and then use the same mask (or a 6mm sphere) during the eigenvariate extraction in the next step, this seems to be doing the same thing two times (i.e., apply mask during results step and masking during eigenvariate extraction step). But, assuming they are independent is there any benefit to applying a mask in the first step (better statistical control?)? Relatedly, if we have an anatomical hypothesis (the 6mm sphere in the dlpfc) and want to use all voxels within that sphere for our PPI, does it then actually really matter what p threshold we use during the results step if we use a mask during the eigenvariate extraction step? Does anyone have any experience with, or advice on, this? Thanks in advance! Kind regards, Anna