Hi Sheng,
You can find all the necessary details regarding your specific questions
in the following reviews:
http://www.ncbi.nlm.nih.gov/pubmed/19914382
http://www.ncbi.nlm.nih.gov/pubmed/20838471
I hope this helps,
Mohamed
On 02/05/2011 21:12, zhang sheng wrote:
> Dear all,
>
> I have some questions about DCM results, and any suggestion is highly
> appreciated.
>
> 1. What does the modulatory parameters (e.g. DCM.a, DCM.b, DCM.c)
> mean? high value means highly correlated and negative value means
> negative correlated?
>
> 2. Let's say we have two DCMs with same structure from two groups data
> (group 1 and group 2), and A->B is one of the modeled connection. If
> we found DCM parameters (DCM.a) of A->B are significant higher in
> group 1 than in group 2. Can we say that the connection of A->B is
> stronger in group 1 than in group 2?
>
> (additional question for this: how to compare DCM parameters between
> two groups? From previous emails, there are two ways. One is to
> directly apply a significant test (e.g. two sample t test), and
> another is to apply a linear contrast to the posterior densities of
> two groups that resulted from using the fixed effects Bayesian
> averaging routine in SPM. However, the results of these two methods
> are totally different. Using Bayesian averaging, we have significant
> difference (p<0.0001). but using directly comparing, we have a very
> non-significant p around 0.69. And when we looked at the DCM
> parameters of two groups. They are 0.1026 vs. -0.0586 from Bayesian
> averaging, but -0.05 vs. -0.06 when we simply got the mean of each
> individual subjects' value. So, which way is correct, Bayesian
> averaging or directly comparing, or did we do something wrong? )
>
>
> 3. Is that possible that we could build a DCM with full connection,
> test the the significance againest 0 of all connections, and then
> simply choose the connections with significant difference with 0 to
> build a best model? If one of the connection say A->B is not
> significant different from 0, can we say that connection A->B is not
> true in the model?
>
> 4. We found something weird in a DCM result. We compared 12 models and
> find the best models are model 9 and 11 using BMS. But, when we looked
> the modulatory parameters, DCM.a matrix, The parameters are all zero
> except self connection parameters. So, what happened here? And We also
> checked the time courses of the VOI, they are highly correlated to
> each other, say around 0.8. Is this the reason?
>
> Thanks,
> Sheng
|