Hi,

I did what you adviced me, but could you check the models that I attached, to see if I was wrong?

I didn't understand the last suggestion about .grp; what does it mean? At the end of the dual_regression's command should I add -e design.grp?

dual_regression groupICA.gica/groupmelodic.ica/melodic_IC 1design.mat design.con 5000 yes <input1>....<input20> -e <design.grp>

Thank you!

Eleonora

> Date: Sun, 5 Feb 2012 13:02:46 -0500
> From: [log in to unmask]
> Subject: Re: [FSL] comparing ICs
> To: [log in to unmask]
>
> Hi,
>
> I think the recommendation would be to put all possible inputs into the same group ICA so that you can obtain one set of spatial maps. Once you have those maps, it's easy to run the dual_regression with a paired t-test design. Just note that you'll have to provide the call to randomise with a design.grp file so that it permutes the correct labels.
>
> Cheers,
> David
>
>
> On Feb 5, 2012, at 12:21 PM, Eleonora Palmaro wrote:
>
> > Hi all,
> > I should compare how ICs modulate before and after an experiment of the same people.
> > Which parameter, considering that there isn't correspondence between the components, could I correlate and how?
> > In some papers it is correlated the component's strenght, if I use the components' value b of Melodic output, will it be ok?
> > Regards
> >
> > EP