Is it possible to apply FDR instead of FWE with randomise?
When I try to change the connectivity from 6 to 26 (--tfce_C 26 ) I get an error
--tfce_C: Missing non-optional argument!
try: randomise --help
Has anyone encountered this?
/Anders
________________________________________
Från: FSL - FMRIB's Software Library [[log in to unmask]] för Mark Jenkinson [[log in to unmask]]
Skickat: den 19 november 2012 11:36
Till: [log in to unmask]
Ämne: Re: [FSL] SV: [FSL] SV: [FSL] Paired t-test with randomise
Or, even easier, make an F-test across the +1 and -1 contrast.
The result is then a two-sided test and you use the standard p-value thresholds.
All the best,
Mark
On 19 Nov 2012, at 16:22, "David V. Smith" <[log in to unmask]<mailto:[log in to unmask]>> wrote:
Yes -- do two contrasts [1] and [-1] and look for voxels exceeding 1-(.05/2) in the corrp map.
David
On Nov 19, 2012, at 11:17 AM, Anders Eklund wrote:
OK thanks, is it possible to get a two sided test somehow?
/Anders
________________________________________
Från: FSL - FMRIB's Software Library [[log in to unmask]<mailto:[log in to unmask]>] för David V. Smith [[log in to unmask]<mailto:[log in to unmask]>]
Skickat: den 19 november 2012 10:59
Till: [log in to unmask]<mailto:[log in to unmask]>
Ämne: Re: [FSL] SV: [FSL] Paired t-test with randomise
Neither of those are used for TFCE, hence the "threshold-free" part of the name. See the documentation and Steve's paper for further details.
Cheers,
David
On Nov 19, 2012, at 10:47 AM, Anders Eklund wrote:
Thanks. If I use threshold free cluster enhancement, is cluster size or cluster mass used?
/Anders
________________________________________
Från: FSL - FMRIB's Software Library [[log in to unmask]<mailto:[log in to unmask]>] för David V. Smith [[log in to unmask]<mailto:[log in to unmask]>]
Skickat: den 19 november 2012 10:06
Till: [log in to unmask]<mailto:[log in to unmask]><mailto:[log in to unmask]>
Ämne: Re: [FSL] Paired t-test with randomise
Hi Anders,
Yes -- I would recommend computing the difference between the two time points with fslmaths (cope_T1 - cope_T2), merging those difference images, and then running a one sample t-test with randomise. If you've already run a group analysis on each time point using FEAT, then you could just use the filtered_func_data files
fslmaths groupT1.gfeat/cope1.feat/filtered_func_data -sub groupT2.gfeat/cope1.feat/filtered_func_data T1minusT2
Hope this helps.
Cheers,
David
On Nov 19, 2012, at 9:37 AM, Anders Eklund wrote:
Dear FSL-experts,
I am analyzing data collected from N subjects at two different timepoints. I wish to investigate, for each voxel, if there is any significant difference in the mean activity between these two timepoints. There is no information about paired t-tests in the randomise user guide, I suppose that the reason is that data from the same subject are not independent and complicates permutatation.
http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/Randomise/UserGuide?highlight=%28%5CbCategoryRandomise%5Cb%29
Is the best solution to calculate the mean difference for each voxel and then simply run a one sample t-test with randomise?
Regards,
Anders
--------------------------------------------
David V. Smith, Ph.D.
Postdoctoral Fellow, Delgado Lab
Department of Psychology
Rutgers University
Newark, NJ 07102
--------------------------------------------
--------------------------------------------
David V. Smith, Ph.D.
Postdoctoral Fellow, Delgado Lab
Department of Psychology
Rutgers University
Newark, NJ 07102
--------------------------------------------
--------------------------------------------
David V. Smith, Ph.D.
Postdoctoral Fellow, Delgado Lab
Department of Psychology
Rutgers University
Newark, NJ 07102
--------------------------------------------
|