Hi,
I'm not sure there is a single right answer here.
The copes will show you the magnitude of the effects, and due to the 4D global intensity scaling, they will have a consistent scaling. However, the z-stats will incorporate information about how reliable those magnitudes were, which could be important. I'd probably try them both. There might be lots of different opinions on this though.
All the best,
Mark
On 13 Aug 2013, at 06:10, Binod Thapa Chhetry <[log in to unmask]> wrote:
> Hi all,
>
> I have analyzed a task based fmri dataset using fsl feat. Now, that I have some interesting finding between two groups in the FEAT 2nd level analysis, I want to take it to the next level : do machine learning(ML) - classification/prediction using PRONTO, or tools similar to that.
>
> My question is, what image(s) per subject from FEAT should I use as my feature/input image for my ML algorithm? Would it be the cope image? z-stat image? or a combination of images on which FEAT performs the ttest during group-level analysis.
>
> For instance, if I wanted to do ML on DTI data, I could use normalized FA images as my feature/input image.
>
> I hope it's not a convoluted problem.
>
> Thanks in advance for the reply.
> Binod
|