Dear Shufei
This sounds like an error - there shouldn't be NaNs in those results. You're welcome to share with me the files needed to reproduce the problem. Alternatively, you could just focus on the other measures - if your objective is to determine which dataset provides the greatest confidence or precision about your model parameters.
Best
Peter
-----Original Message-----
From: SUBSCRIBE FSL Shufei Zhang <[log in to unmask]>
Sent: 24 October 2022 13:23
To: [log in to unmask]; Zeidman, Peter <[log in to unmask]>
Subject: Re: spm_dcm_bdc - Bayesian data comparison
⚠ Caution: External sender
Dear Peter,
We are using spm_dcm_bdc.m to validate our comparisons between two datasets. What we can observe is that we found differences in measures (parameter certainty and information gain from parameters), however, the information gain from models is NaN. There are no errors reported during running. I am not sure that it (NaN) is normal.
Introduce our models shortly. We used the CSD-DCM for both datasets and a full-connected model (Mat-A) was applied for them.
Thank you for your time in advance. Any help would be appreciated.
Best regards and all the best,
Shufei Zhang
|