Hi Vadim,

Please, see below:

On 18 July 2016 at 17:16, Vadim Axelrod <[log in to unmask]> wrote:

Dear list,

I need to run random-effects group analysis (i.e., comparison between two groups) and I am confused about my options in FSL.

Group analyses can also use fixed-effects. In any case, let's focus on random effects that it seems you need.

On one hand, in FEAT there is a "high-level analysis" and "inputs as 3D cope images from FEAT" options. To use this FEAT pathway, I need my images to be registered in first-level analysis (which is not my case). In addition, FEAT calls internally FLAMEO, so I can run directly FLAMEO and probably there is no requirement to have the images registered (https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=FSL;b447798e.0909).

To compare groups of subjects, the subjects will have to be registered to a common space. That is, registration cannot be skipped.

In addition, there are Randomise and PALM. But are they suitable for random-effects group analysis?

Both randomise and PALM do fixed-effect analyses.

In general, I analyze HCP connectome surface data, so if the analysis can be done directly with surface data, it would be preferable. As far as I understand PALM does support surface data.

Yes, PALM supports surface and should be able to read/write CIFTI, provided that the Connectome Workbench is installed. However, fixed effects only (this will change soon but I'm reluctant to give a timeframe).

All the best,

Anderson

Many thanks for you help,
Vadim