Dear Kushal Kapse,
Here I would advise using Bayesian Model Averaging to first average over model space.
This will produce a single vector of model parameters for each subject:
w=BMS.DCM.rfx.bma.mEp
You can then do t-tests, F-tests or multivariate tests to look for consistency of
effects over a group of patients or eg. two-sample t-tests to look for differences between patients and controls.
Best wishes,
Will.
> -----Original Message-----
> From: SPM (Statistical Parametric Mapping) [mailto:[log in to unmask]]
> On Behalf Of Kushal Kapse
> Sent: 05 September 2013 17:16
> To: [log in to unmask]
> Subject: [SPM] BPA, BMA v/s FFX, RFX
>
> Hi Peter and SPM users,
>
> Two months ago we had email conversation about my doubts on DCM. I
> finished the analysis and now working on interpretation. I went through
> the DCM tutorial and BMA paper to get understanding on what way to
> interpret my datasets. I read the book section
> http://www.fil.ion.ucl.ac.uk/~wpenny/publications/spm-
> book/selection.pdf to understand BMA and BPA. But i have following
> question:
>
> 1- We are specifically looking to understand connectivity in brain
> across treatment. This makes us incline to use parameter estimates over
> model estimates (BPA over BMA). BUt seems we can only do BPA if we
> select FFX and BMA if we select RFX. Why is this? Can I do RFX and then
> perform BPA? The reason to use RFX is that we are expecting different
> cognitive network for each patient. But we want to do Bayesian
> Parameter Averaging to interpret how patients connectivity changes as a
> measure of treatment across two scans.
>
> SO my question is , can i do BPA using RFX.
>
> Your advice will be really helpful again.
>
> THanks
>
>
|