 |
 |
Dear Hilda, To consider how best to analyse your data, let's get a bit more detail: 1. Is this resting state or task based? 2. Are you comparing between groups of participants or comparing between tasks within participants? 3. Let's say you could accurately estimate an 11 ROI model with all possible connections in each subject. That's 110 between-region connections. How would you draw conclusions from this result? Do you have any ideas which connections should show an effect, or would any result be equally surprising? Or are you interested in properties of the whole network (graph theory measures?) . Best, Peter. From: Hilda Azimi [mailto:[log in to unmask]] Sent: 07 July 2014 20:06 To: Zeidman, Peter Subject: Re: Bayesian model compression Dear Peter, Unfortunately I don't have any special hypothesis to reduce the connections.Is there any other way to use for group analysis? On Monday, July 7, 2014, Zeidman, Peter <[log in to unmask]<mailto:[log in to unmask]>> wrote: Dear Hilda, You may be able to use post-hoc DCM for this. For each subject, you estimate a 'full' model and it will search through reduced models. However, it assumes the full model could be correctly estimated, and this probably won't be possible if you allow all connections between all 11 ROIS. Can you use your hypotheses about the connectivity to select just certain connections between these regions? Best, Peter. From: SPM (Statistical Parametric Mapping) [mailto:[log in to unmask]<javascript:_e(%7B%7D,'cvml',[log in to unmask]);>] On Behalf Of Hilda Azimi Sent: 05 July 2014 13:17 To: [log in to unmask]<javascript:_e(%7B%7D,'cvml',[log in to unmask]);> Subject: [SPM] Bayesian model compression Dear Experts; I want to use "Bayesian model compression" as group analysis.But i have 11 ROIs for each subject which result in too many models.I wonder if there is any solution to solve this problem and specially the one related to optimization and post-hoc code? Thanks; Hilda