Dear Gerald,
On Thu, Nov 5, 2009 at 1:10 PM, Gerald Cooray <[log in to unmask]> wrote:
> I will be comparing 2 groups of subjects. There is only 1 type of trial
> per subject. My plan is to run 1 model of DCM on all subjects and then
> use ANOVA to compare the data regarding the A and C matrix estimates
> between the 2 groups. I have entered into the script that there is only
> 1 trial.
>
>
> Can I put the B matrix to zero since there is no comparison between
> trials, or will this perhaps mess the program. I have tried and it seems
> to work, although I get a quicker convergence than if B is nonzero.
What I'd recommend you to do is to set up a DCM once with GUI, save
it, look what the struct looks like and then either build the same
struct with a script or just load that struct and modify some specific
fields (like the input file) with a script. If you have only one trial
type there will be no B matrix in the GUI. See what the resulting
struct looks like and do the same (I just don't know).
>
> Should one use the hanning window for ERP signals, I remember you
> mentioned that during the SPM course.
>
My philosophy is to use as few pre-processing tricks as necessary to
see the effect. Hanning window basically suppresses baseline
fluctuations and late part of the response and lets you focus on the
central part of it. If you have a lot of those things perhaps you
should use it, but if you can get good results without, perhaps you
shouldn't.
Best,
Vladimir
|