Print

Print
Hi - note that the design.grp file means different things to flameo (group stats in FEAT) vs randomise.

Apart from that, there are several differences between flameo and randomise.  The former has better modelling (Bayesian mixed effects) but the latter probably has more accurate control of FWE FPR in the presence of smoothness (because the former relies on GRF).

Cheers.



On 14 Dec 2011, at 10:16, Maria wrote:

Dear FSL wizards.

I'm trying to run randomise with TFCE on my higher-level repeated measures (N15) filtered_func image, but not getting any results at all. Tried with different amounts of spatial smoothing (1.5 vs 5 mm), but still nothing.

I have been able to find significant clusters using cluster thresholding in higher-level FEAT. However, had to fiddle around a bit with the threshold, so feels a bit arbitrary (hence my wanting to try tfce).

Also tried cluster thresholding with randomise (with both amounts of smoothing and several different thresholds), but again, nothing.

Might I be doing something wrong here?

I am using the mask made by FEAT and have changed the design.grp file to account for repeated measures.

Thanks a bunch,
Maria



---------------------------------------------------------------------------
Stephen M. Smith, Professor of Biomedical Engineering
Associate Director,  Oxford University FMRIB Centre

FMRIB, JR Hospital, Headington, Oxford  OX3 9DU, UK
+44 (0) 1865 222726  (fax 222717)
[log in to unmask]    http://www.fmrib.ox.ac.uk/~steve
---------------------------------------------------------------------------