Dear SPM-ers,
I am planning a series of analyses with the aim of investigating similarities and differences in effective connectivity matrices across ~4 groups (healthy controls, depression, early dementia, and established dementia).
The plan is to use spectral DCM (on resting state fMRI data) per previous work by Razi and Friston et al., and I was wondering if there is a principled approach to examine for similarities between networks. I have previously used the Network Based Statistic (Zalesky et al.) to examine *differences* in optimised DCMs across groups of depressed subjects, but my hypotheses for the current study are that networks will be similar across depression and dementia.
I saw a recent paper from your group that used Pearson correlations to test for similarities between resting and task based DCM, but I wasn't sure if there was a more principled approach that could be recommended? (i.e., is there a statistical approach for finding common networks across groups)?
Broadly, would Bayesian Model Comparison be useful to test the hypothesis that optimised DCMs are more or less consistent across groups?
Thanks in advance.
Matt
|