Hi,
this has been on the list quite a few times.
To quote Christian Beckmann:
"The first column is the rank-1 time course for this particular component (what's being plotted in the graph) and the remaining 20 column represent how this effect looks temporally in each of your 20 subjects."
…and Stephen Smith:
"[…] the "representative" timecourse is not very meaningful (it's created by a PCA across all sub-timecourses combined somehow) - this is why to get meaningful indivdual subject timecourses you need to use dual-regression as a next step."
and Christian again:
"In general, the GLM output is based on a post-hoc rank 1 approximation of all time courses to obtain a single time course and corresponding subject mode vectors. If you don't specify design matrices then the only GLM test being performed is a 'single group average' design at the between subject level, i.e. melodic tests if the subject mode
vector is >0."
Browse the list for "melodic output" etc.
Best regards
Cornelius
Am 10.07.2012 um 10:00 schrieb Emma Biggs:
> Hi,
>
> I was wondering if anyone could help - I have some questions about the output (report) from MELODIC (concat-ICA)
>
> First, for each IC there is a graph showing the time-course of the component (Temporal mode): this covers 360s, which makes sense (180 time points x TR 2s), except that I can't figure out where the time-course comes from? Is it the time-course of one subject (and if so, which one?) or an average time-course? When I click the graph I get a text file with 7 columns (360 lines)...but I input 6 subjects, and none of these columns is an average of the other six...
>
> Second, the sessions/subjects mode: I didn't input a design matrix at this stage (I went for the dual-regression approach) but there are F and p values reported. Do these refer to the fit of the mixture model? And the contrasts to the gaussian model and 2 gamma model?
>
> Third, at the top of the screen for each component the % of explained variance and % of total variance are reported. What variance do these refer to?
>
> Thanks for your help!
>
> Best wishes,
> Emma
|