Peter, thank you for all your help! On Tue, May 19, 2020 at 2:15 PM Zeidman, Peter <[log in to unmask]> wrote: > Dear Tammar > > > We would like to ask your opinion on two matters, please: > > 1. We have 3 conditions in our study (Experimental, Control, Baseline). > Our DCM includes only Experimental & Control conditions. > We compared two analyses: > > i. DCM using F contrast that includes all three conditions. > > ii. DCM using F contrast that includes only two conditions (without > Baseline) We found a difference in the explained variance. > > It seems weird because the Baseline condition is not defined in the > intrinsic conditions in both DCMs. > > What is the reason? Which F contrast is recommended? > > The F-contrast tells SPM which effects to regress out of the time series > during VOI extraction. Are there unmodelled rest periods or inter-trial > intervals in addition to the Baseline condition? If so, I recommend you > keep baseline in the F-contrast (i.e., don't regress it out). If there are > no rest periods - so the entire time series is Experimental, Control or > Baseline, then I don't think it really matters either way. Use whichever > gives you nicer results. (My personal preference is to only regress out > factors like breathing, head motion, session means.) > > > 2. Is there any preventive to include many between-regions modulatory > connections? > > We did not find so, but we want to make sure it is doesn't come at the > expense of something else. > > Use as many parameters as you think might be necessary to explain your > data. Then assess whether you could have used a simpler model, by using > Bayesian model reduction to switch off particular parameters. > > Best > Peter > -- תמר טרוזמן קלינאית תקשורת, MA מ.ר. 13-138103