Dear all,
I have some questions regarding model averaging for resting state data using spectral DCM analyses.
I have two subject groups, and I tested the default mode network using the 30 possible DCM models constructed from the Di and Biswal (2014) paper.
Using the random effects analysis, the winning model for group 1 is 5, whereas the winning model for group 2 is between 5 and 10 (both posterior probabilities are very close, 0.297 and 0.273).
In order to compare the two groups, I have adopted two methods.
First, I extracted the connection strengths for the winning model 5 in both groups for comparison.
Second, since the winning model for group 2 is 5 and 10, I performed Bayesian model averaging for each group. However, I noticed that the winning models selected for each subject for Bayesian model averaging purpose may be different. Did I do this correctly? I subsequently found the mean and standard devidation of each of connection strengths across subjects for each group. How do I then test the differences between groups?
Thanks for your help!