Hi Christian,
 
Thank you for your replay.
We still have the same problem. The thing is that we want to evaluate different subjects design on the same components. The option that you proposed follows concatenating the subjects, it doesn't give any output for the subjects contrast.
 
Thus far, what we did, was first to decompose the data and get some IC's of interest, and then, we evaluated these ICs on different GLM subject's designs. The problem is that spatial maps between runs are slighlty different (although visually we know are the same), and we don't know how can we assure that. That's the reason why we were trying this mixture modelling.
 
Do you have other suggestions? Or do you think that we can assume directly that the components are the same between runs?
 
Thanks in advance for your response,
Eva 
 

2011/4/14 Christian F. Beckmann <[log in to unmask]>
Hi Eva,

If you only want to run post-stats then the input file is only used to supply melodic with the image which is used as the background image. Therefore, just use the background image in the .gica folder and things should be fine (or use the MNI template if that is what you want to use as the bg)

melodic -i ${FSLDIR}/data/standard/MNI152_T1_2mm --ICs=melodic_IC --mix=melodic_mix [other options]

should do the trick. Just make sure that the input file has the same spatial resolution as the component file (e.g. if you downsampled to 4mm isotropic you need to use a 4mm iso template brain or upsample melodic_IC to 2mm isotropic resolution).
hth
Christian



On 13 Apr 2011, at 15:45, Eva Palacios wrote:

> Hello,
>
> we would like to run only the post-stats part of a melodic analysis, using a set of already obtained ICs. Is it possible to run mixture modelling on this maps indicating a Subjects design and Subjects contrast matrices? We have tried already this option and it seems that melodic is concatenating timeseries of all the subjects, we used this command:
>
> melodic -i .filelist - -ICs=groupmelodic.ica/melodic_IC.nii.gz - -mix=groupmelodic.ica/melodic_mix --Sdes=../design.mat --Scon=../design.con --report -o outdir -a tica -v
> When doing the standard procedure (putting Sdes inside Melodic GUI) we realized that spatial maps of the components are slightly different between analyses.
>
> thank you
>
> Eva