Hi, Matthew
Thank you very much for your information.
I have a few further questions regarding this and would really appreciate if you would shed some light on.
Since my data were derived from TBSS and the major purposes are to investigate 1) hemisphere effect, 2) handedness effect and 3) interaction of the two, I am wondering if it is correct for me to use the following statistics to check what I want to see:
Data preparations:
1) Original FA maps (36 subjects) and their flipped FA maps were fed into TBSS processing pipelines. A symmetrical FA template was used for the FA images to register on.
2) Then, the right hemisphere of the derived 4D FA (totally 72 volumes) skeleton was masked out and the statistics were carried on the left half, the first 36 volumes represented the FA images on the right hemisphere (from the flipped images) and the next 36 volumes represent original FA images.
Statistics:
1) Handedness measures were continuous ranging from [-1, 1]. However,
correlation results did not show significance between handedness index
and asymmetry quotient of FA on a voxel basis ( AQ = (FA_L –
FA_R)/(FA_L + FA_R). therefore, I divided the subjects into two groups
(based on their handedness and conducted group analysis).
2) One sample paired t-test on FA difference map (Left hemisphere subtracts right hemisphere) was used to check hemisphere effect;
3) Two sample unpaired t-test with age as nuisance covariate on FA map to check handedness effect;
4) Two sample unpaired t-test with age as nuisance covariate on FA difference map to check hemisphere-by-handedness effect;
Are these tests correctly set up to let me see what I want to check? If not, is there a more efficient way to set up the design?
Many thanks in advance
Longchuan
________________________________
From: Matthew Webster <[log in to unmask]>
To: [log in to unmask]
Sent: Tuesday, May 19, 2009 11:29:49 AM
Subject: Re: [FSL] 2-by-2 ANOVA using randomise
Hi,
Randomise can handle the within-subject factor by using the -e option. It currently is unable to deal with between-subject factors for this kind of analysis, although we are intending to enable this feature in future versions.
Many Regards
Matthew
> From: Long Chuan <[log in to unmask]>
> Date: 14 May 2009 21:56:35 BST
> To: [log in to unmask]
> Subject: [FSL] 2-by-2 ANOVA using randomise
> Reply-To: FSL - FMRIB's Software Library <[log in to unmask]>
>
> Hi, FSL experts
>
> I think I have posted similar questions before but this question is still confusing me.
> I have run TBSS registration on a group of subject. I measured the handedness of the subjects. What I am interested in seeing are 1) hemispheric effect, 2) handedness effect and 3) the interaction of the two. By searching the old archives, it seems to me that randomise in this version cannot take care of 2-by-2 repeated ANOVA (between subject factor of handedness and within-subject factor of hemisphere, also age is used as covariate).
>
> So my question is:
> 1) Is there any way that I can model this 2-by-2 ANOVA design in randomise and get the results in a single run? If not, what is the most proper way to carry this analysis in randomise?
>
> Any suggestions or advice is greatly appreciated
>
> Longchuan
