Dear Gregor,

On Wed, Nov 28, 2012 at 5:26 PM, Gregor Strobbe <[log in to unmask]> wrote:
> Dear SPM experts,
>
> I would like to ask your advice on the statistical analysis of EEG source
> reconstructed images.
>
> I have the following data:
>
> Grand averaged ERP data (over subjects)  for P300 and MMN paradigms in 3
> different conditions, for both women and men.
>
> A number of generated nifti images in different time windows corresponding
> with the grand average ERPs.
>
> (I merged all the ERPs and generated the nifti images all at once for
> further analysis)
>
> 3 questions:
>
> 1. Is it justifiable to do a within "subject" ANOVA assuming the different
> "subjects" in the analysis as the nifti images corresponding with different
> time windows? If yes, can the time windows overlap?

Since you need some reasonable number of samples to do an ANOVA I
don't see how you can get that from a single ERP. It sounds like you
want to do multiple sliding windows in different parts of the waveform
and that would violate the assumption of the ANOVA that the samples at
each level are independent. You would have quite a complicated
dependency structure not accounted for, as if in the extreme case you
just took each sample many times without informing the model about it.

>
> 2. Or would you advice to do the source reconstruction for each subject
> separately and use the corresponding generated images in the ANOVA?

Yes, that'd be the standard strategy. You could use the group
inversion option to ensure better spatial overlap between activations.

>
> 3. Is it actually possible to do within subject statistical analysis, for
> example based on the the activity in different labeled regions if there's
> only 1 subject available?

You could export an image per trial and then do statistics across
trials as those would be truly independent.

Vladimir

>
> Thanks a lot for the answers
>
> Grtz
>
> Gregor
>
>
>
> --
>
> Gregor Strobbe
>
> Doctoral researcher
> IBiTech - Medisip (Ghent University)
> De Pintelaan 185, Blok B
> BE-9000 Gent (Belgium)
> Tel: +3293324322
> http://medisip.elis.ugent.be/
>