Dear experts,
considering the longitudinal design of two groups, would it also make sense to set up the randomise-design in an "ANCOVA-fashion", i.e. to feed the 4D data from timepoint 2 into randomise, model the interaction effect between time + group by a two-sample t-test, and to add the 4D data from timepoint 1 as a voxelwise regressor (covariate)? As far as I understand, looking at longitudinal designs using repeated measures approaches (~paired t-tests) or using ANCOVA approaches may result in slightly different statistics. But actually I am not sure whether randomise can handle voxelwise regressors..
Many thanks for your advice!
best, Veronica
________________________________________
Von: FSL - FMRIB's Software Library [[log in to unmask]] im Auftrag von Gwenaëlle DOUAUD [[log in to unmask]]
Gesendet: Donnerstag, 16. Februar 2012 13:53
An: [log in to unmask]
Betreff: [FSL] Re : [FSL] Sanity Check: Longitudinal TBSS analyses
Hey Kirstie,
just to add to what Steve said earlier, you might want to deproject (in the MNI space) your results for all the subjects, so that you can be confident that you are comparing the same tracts, especially within subject... Also, you can just run directly fslmaths -sub on your all_FA_skeletonised_T1 and all_FA_skeletonised_T2 (both obtained with the -T option for tbss_3_postreg) to get your substracted 4D image for randomise.
Cheers,
Gwenaëlle
--------------------------------------------------------------------
Gwenaëlle Douaud, PhD
FMRIB Centre, University of Oxford
John Radcliffe Hospital, Headington
OX3 9DU Oxford UK
Tel: +44 (0) 1865 222 523
Fax: +44 (0) 1865 222 717
www.fmrib.ox.ac.uk/~douaud
--------------------------------------------------------------------
________________________________
De : Kirstie Whitaker <[log in to unmask]>
À : [log in to unmask]
Envoyé le : Mercredi 15 février 2012 23h24
Objet : [FSL] Sanity Check: Longitudinal TBSS analyses
Hi FSL gurus,
I'm running an analysis looking at the relationship between changes in white matter (time 1 to time 2) for two different groups.
In response to this email<https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=fsl;d27f897d.1111> from November 2011, Tom Nichols advises:
--
Sorry to confuse things, but as Donald just pointed out to me off-list, you have a 2x2 mixed design and so what I say below (regarding a paired t-test) is irrelevant. Donald is correct, since you have one *between* subject factor (in addition to a within subject factor) randomise cannot fit this data with this model.
The only way to analyze this presently is fit three separate models, as Donald and others frequently prescribe: Average the pairs, and fit a two-sample t-test to get main effect of group; difference the pairs, and fit a one-sample t-test to get the main effect of baseline/follow up; and fit the difference data with a two-sample t-test to get the interaction.
--
I've run all participants through the tbss processing steps and, to create the appropriate difference images, I've subtracted the *skeletonised* data for T1 from T2. My question is whether this is the best way to subtract these two images. I tried subtracting the two images before they were skeletonised (but in MNI space) and was then unsure of how to skeletonise these maps. I also figured that subtracting the skeletonised data meant that I'm testing the same values that would be put into a unpaired ttest.
Any thoughts on the best way to subtract (or average) T1 and T2 data would be really helpful. If it's relevant, I'm also planning to use these difference maps to relate changes in white matter to changes in behavior.
Thank you!
Kx
--
It's that time again! I'm riding from San Francisco to LA with AIDS/Lifecycle for the fourth time in June 2012.
Please donate anything you can to help me reach my goal: http://www.tofighthiv.org/goto/kirstie
-----------------------
Kirstie Whitaker
Doctoral Candidate
Cognitive Control and Development Laboratory
Helen Wills Neuroscience Institute
University of California at Berkeley
134 Barker Hall, MC 3190
Berkeley, CA, 94707
tel: 510 684 2456
web: bungelab.berkeley.edu<http://bungelab.berkeley.edu/>
|