We’ve specified a gPPI analysis in CONN, that uses the output from tedana that’s been normalised to 2mm MNI space. Each subject (N=35) has 8 runs, and for a “normal” fMRI analysis (in FSL) we’d combine these runs using a fixed effects model per subject, then take the 35 outputs into a mixed effects analysis.
Does CONN do this for your automagically? We can see how to run a 2nd level analysis, but are not sure if CONN is doing the heavy lifting behind the scenes.
Just as a follow up - looking at the tutorials for running gPPI analyses with CONN it appears that it has the same limitations as running with FSL i.e. it doesn’t try to deconvolve the neuronal activity from the seed region (physiological signal):
Hi Jon and Eszter, I’d keep the eight tedana outputs as separate sessions in CONN rather than concatenating them, with each run’s timing and nuisance regressors entered separately. CONN should then combine the session-level information into subject-level gPPI contrasts for the second-level analysis, so you likely do not need a separate FSL fixed-effects step. You’re also right that CONN does not use the same neuronal deconvolution approach as SPM. If that is essential for your analysis, the SPM gPPI toolbox may be the better route.
