Hi Stefano,
> 1) User guide suggests to specifiy a single regressor and to perform a single f-contrast with the --fonly option. Is it possible to perform an ANOVA and interpreting results using t-contrasts? I find a bit tricky to use the suggested timeseries option to define the direction of change.
That simple model is really just an example - you can use any model that can be run with randomise. See http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/GLM for more examples. In particular, the 1-factor 4-levels ANOVA example sounds similar to what you want (except for the number of levels obviously).
> 2) User guide also suggest to use cluster-based mutiple comparison correction. In your experience, is there any limitation to TFCE use?
TFCE should work just fine. Make sure that you use the 2D optimised version, i.e. the --T2 option in randomise.
Hope that helps!
Cheers,
Eelke
|