Hi
> In the web manual (http://www.fmrib.ox.ac.uk/fsl/melodic/index.html) :
> - melodic_mix should contain subject temporal response for each component concatenated.
> - melodic_Tmodes should contain subject temporal response for each component + first eigenvector of each component.
>
> Subject temporal response does not match between this two files (even with a scale factor)
>
I'd be surprised if this was true, have you looked at the correct entries in one file relative to the other?
> Could anyone explain what are differences between melodic_mix and melodic_Tmodes ?
>
apart from scaling and presence of first Eigenvector in melodic_Tmodes the two files only differ in the order. If N_t is #time points, N_c is #components and N_s is #subjects in your analysis then
melodic_mix is a (N_t x N_s) x N_c matrix, each column contains the different temporal responses stacked vertically, i.e. temporal response for subject 3 in component 2 is melodic_mix((N_s-1)*N_t+1:N_2*N_t,N_c)
melodic_Tmodes is a N_t x ((1+N_s)*N_c) matrix, each column containing a single temporal response, i.e. temporal response for subject 3 in component 2 is melodic_Tmodes(:,(N_s+1)*(N_c-1)+1+N_s)
These time courses then only differ in scale.
hth
Christian
|