Hi,
> 1- Since my data are from different studies, subjects have different TRs (three different TRs = 2, 2.04, 2.3, around 20 subjects per group) is there a way two run all of them together, what should I put for my TR in MELODIC?
>
You can feed all data in, the TR setting in the GUI is not being used, other than convert the x-labelling in the report to be in terms of s rather than volumes.
> 2- Each subject has two scan drug-placebo. Is this the correct way to compare groups:
> - Run the Multi-session temporal concatenation in MELODIC for each condition (drug,placebo) separately
> - Use GLM to compare them
> - Use dual-regression to map the networks to individual subjects
The steps are
(i) derive template maps (e.g. using group ICA on your data)
(ii) run DR to map these to individual subjects
(iii) use randomise to test for differences in the DR derived maps.
Any difference in (iii) will depend on the initial templates (i) and any bias in the template definition, i.e. when you derive templates only from placebo data and then use this to derive the stats of drug vs placebo then the inital templates will be 'closer' to the placebo data than to the drug data so the analysis becomes biased. For paired deisgns this is less of an issue but for nonpaired designs this is to be avoided. One option is to run the initial group ICA on the full data (all data sets combined) which - under the null assumption of no difference between placebo and drug - is unbiased.
> or there is another way to do it.
>
If you happen to have a second control session then you could use group ICA across those sessions to derive an unbiased template set of maps.
hth
Christian
> Thanks in advance,
> Yashar
|