Yingying

>>> Michael Harms <[log in to unmask]> 10/25/2011 12:22 PM >>>
Hi Yingying,
If you read on to the next paragraph in the randomise documentation,
you'll see a description of the approach that randomise uses when
nuisance covariates are present. So, it is NOT the case that randomise
can't be used in the presence of nuisance covariates. Rather
permutation testing becomes more complicated because it is no longer
simply a matter of simply exchanging group labels. Randomise takes care
of all that internally.

cheers,
-MH

On Tue, 2011-10-25 at 12:02 -0400, Yingying Wang wrote:
> Hi Josselin,
>
> design mat:
>
> 1 49
> 1 34
> 1 12
> ....
> 1 12
> 1 22
> 1 23
>
> How do you want to look at the age effect? Do you have age group?
> Randomise is a non-parametric testing. It might not work well with
> age effect. Look at the this paragraph on the website:
> Use of almost any other GLM will result in approximately exact
> inference. In particular, when the model includes both the effect
> tested (e.g., difference in FA between two groups) and nuisance
> variables (e.g., age), exact tests are not generally available.
> Permutation tests rely on an assumption of exchangeability; with the
> models above, the null hypothesis implies complete exchangeability of
> the observations. When there are nuisance effects, however, the null
> hypothesis no longer assures the exchangeability of the data (e.g.
> even when the null hypothesis of no FA difference is true, age effects
> imply that you can't permute the data without altering the structure
> of the data).
>
> Use fsl_glm or use 3dRegAna, or SPM ... other tools for the
> statistics. According to my experience with randomise, it works well
> for simple study design (not for fancy or complex one)
>
> Cheers,
> Yingying
>
> >>> Josselin Houenou <[log in to unmask]> 10/24/2011 12:51 PM >>>
> Hi,
>
> Thanks for your answer. I tried randomise in another sample of 80
> subjects. As you said, I indeed found a significantly lower FA in
> females than in males, using the --T2 option and TFCE.
> Nevertheless, when I look at age effects in a sample, I don't manage
> to find any effect, even when I lower my p value to 0.75.
>
> I am doing:
>
> randomise -i all_FA_skeletonised -o tbss -m mean_FA_skeleton_mask -d
> design.mat -t design.con -n 500 -D --T2 -V
>
> tbss_fill tbss_tfce_corrp_tstat4 0.75 mean_FA tbss_fillfslview mean_FA
> -b 0,0.6 mean_FA_skeleton -l Green -b 0.2,0.7 tbss_fill -l Red-Yellow
>
> fslview mean_FA -b 0,0.6 mean_FA_skeleton -l Green -b 0.2,0.7
> tbss_fill -l Red-Yellow
>
>
>
> It really questions me about the sensitivity of TBSS analyses if
> anyone get an idea ?
>
> Thanks,
> Josselin
>