Dear Takuya
You are correct that the results produced by PEB reflect a Bayesian model average. In more detail, all subjects' connectivity parameters are taken to the group level, and modelled using a general linear model (GLM). The GLM has one regressor per connection per covariate. Multiple GLMs are then compared, with different mixture of connection parameters switched on and off, to identify which mixture gives the highest free energy. The parameters of the best GLMs are then averaged to produce the bar charts in the output (BMA).
The best parameters that describe the group are not necessarily the best parameters to describe any one individual. Therefore, I suggest you simply use the parameters from each subject's DCM to produce your scatter plot - i.e., the parameters that were estimated before conducting the PEB analysis. Let us know if you need help scripting that.
Best
Peter
-----Original Message-----
From: SPM (Statistical Parametric Mapping) <[log in to unmask]> On Behalf Of Takuya Ishida
Sent: 16 April 2022 15:52
To: [log in to unmask]
Subject: [SPM] DCM:how to extract each subject’s dcm parameters after Bayesian model averaging of srcond analysis
⚠ Caution: External sender
Dear experts
I am performing the second analysis of spectral DCM analysis based on the parametric
empirical Bayes (PEB) framework.
When I know the association between some specific DCM parameter (connectivity value) and
some clinical score, I want to perform the scatterplot of those two parameters.
But when I know that association using the spm_dcm_peb_review function after
performing spm_dcm_peb_bmc(PEB), I think each DCM parameter (effective connectivity value) was
calculated by performing Bayesian Model Averaging.
I already specified each subject's DCM parameters using the fully-connected model
as a first-level analysis.
But PEB adopted the DCM parameter which was averaged across the 256 pruned models
from the fully-connected model to look at the association between each DCM parameter
(connectivity value) and the clinical score.
So I want to extract those Bayesian-Model-Averaged DCM parameters of each subject
after performing PEB analysis.
Could anyone help me extract them?
Any comment is appreciated. I am sorry if my question is something irrelevant.
Thank you for your help.
Thank you.
Best
Takuya
|