Christian,
So a component is a volume, not an event?
Emily
----- Original Message -----
From: "Christian F. Beckmann" <[log in to unmask]>
Date: Saturday, June 14, 2008 7:03 am
Subject: Re: [FSL] ICA components
> Hi
>
> Firstly, if your run length is 180 volumes and you specify 180
> components then this won't work and melodic will resort to
> estimating
> the components instead.
>
> Secondly, for correlation analysis it's easiest to use the Glm gui
> to
> set up a FEAT design matrix and use that as your timeseries model.
> hth
> Christian
>
>
>
>
> On 12 Jun 2008, at 21:24, Emily Stoneham wrote:
>
> > Hello Again!
> >
> > I know no one has gotten back to me yet on my previous question
>
> > about identifying ICA
> > components since it was just posted, but I wanted to find out if
> my
> > workaround for the identification
> > is a legitimate approach, and to find out if there is a way to
> hand
> > label them based on what I found:
> >
> > I ran Melodic, and specified exactly how many components I knew
> we
> > 'should' see. I then took the
> > behavioral data the feat for that data set was based on, and
> > organized it in time, labeling each
> > progressive point (expected components) so that I could
> correlate
> > them to the labeled components
> > from ICA.
> >
> > Is this a valid way to do this?
> >
> > Regards,
> > Emily
>
|