Dear Lau,
I think the first question that you should ask yourself is what it is
you want to get as an input for your classifier. If you want to get
something that looks like a functional image of the brain that matches
the structural image you should use 'M/EEG inversion results' batch to
make images rather than extract. Extract gives you source waveforms
for some ROIs you specify. It doesn't matter whether you input your
data one by one one or many datasets at once, the results of extract
will be the same. It does matter only for the inversion module where
you will get a series of individual inversions or a group inversion
respectively. If you then want to do statistics on the extracted
waveforms, you can do either comparisons in 1D or do time-frequency
analysis (for that you would normally need an epoched dataset to begin
with). 1D analysis is not well supported in SPM8. There is an example
script I posted on the list before and I can post again if that's what
you really need. Time-frequency analysis and stats are pretty standard
and are described in the manual.
Best,
Vladimir
On Fri, Feb 1, 2013 at 9:09 AM, Lau Møller Andersen
<[log in to unmask]> wrote:
> Dear SPM-list
>
> I have some questions regarding source extraction as it is done in the batch in SPM8.
>
> I have run a visual experiment with 4 conditions. Ultimately I want to train a classifier and see if it can distinguish between the 4 conditions based on the field intensities of extracted sources.
>
> The first step I have taken is to collapse all 4 conditions and find the most responsive sources. Then I created a source extraction batch that I ran on the uncollapsed files, which thus had the 4 conditions labelled. Here's my first question. Should I run one batch for each subject, i.e. create a source extraction file for each subject, or should I create one, which takes all my subject files as input? I am not really sure what the latter does?
>
> The second question regards how I would be able to make statistical tests contrasting the four conditions. I expected I could create images from the source extraction files, but I get the error message that the fiducials are missing. How can I supply those?
>
> I apologise if my conceptual understanding of what the source extraction command does is lacking, but I have had a hard time finding a description of its exact function anywhere in the documentation.
>
> Best Wishes
>
> Lau
|