Hi Jo,
It looks like you are doing the right things.
Have you checked that the registrations are all good?
For instance, have you looked at all_FA and do the
individual subjects look well aligned?
If they do then I would check the values in one of
the voxels that is showing the effect and see what
the FA values are like in the different subjects. This
is the best way to see whether they are actually going
the way around that you think they are or not. Don't
forget that FA can increase when a crossing fibre
population is being reduced.
Anyway, try that and hopefully things will get a little
clearer.
All the best,
Mark
On 2 Jul 2010, at 04:33, Joanne Lin wrote:
> Hello,
>
> Thanks for that J I was careful to enter my data into the matrix in
> the same order as my original FA files. And I also tried the
> ‘design_ttest2 design 21 16’ command with the same results so
> hopefully someone will be able to shed some light. Otherwise I’ll
> be trying to explain some very interesting results!
>
> Cheers,
> Jo
>
> From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On
> Behalf Of haiyan zhang
> Sent: Friday, 2 July 2010 3:24 p.m.
> To: [log in to unmask]
> Subject: Re: [FSL] TBSS two-group unpaired t-test
>
> Helllo. You data must have the same order with the matrix.
> Because I am also learn to use TBSS now . I don't know much.By the
> way ,I find they all use the command "design _ttest2 design 21 16" I
> don't know whether there is any differents between the command and
> the GUI.
> Good luck!
> .
> 2010/7/1 Joanne Lin <[log in to unmask]>
> Hi all,
>
> I am running my first TBSS analysis comparing FA of 2 groups
> (controls v
> methamphetamine addicts) - not controlling for any covariates at this
> stage.
>
> I've been closely following the TBSS manual, lecture notes from the
> FSL
> course (super duper excellent!) and practicals and had a look through
> the archives on unpaired t-tests so I hope I've done this right.
>
> My design matrix is basic - 2 groups (21 controls v 16 addicts), 2 EVs
>
> EV1 EV2
> 1 0
> 1 0
> 1 0
> 0 1
> 0 1
> 0 1
>
> And so on...for the correct number of participants, following
> alphabetical order of my original FA files.
>
> I have set up 2 contrasts to look for differences in either direction
> (attached image file) -
>
> 1 -1 (Control > Meth)
> -1 1 (Meth > Control)
>
> Then I ran randomise as written in part 5 TBSS.
>
> Can anyone spot anything I've done wrong with the matrix and/or
> contrasts? Somewhere where I could have got things the wrong way
> around? Because the results were not what I expected at all - the
> opposite in fact.
>
> I really hope I made a mistake because it's strange that
> methamphetamine
> addicts have higher FA than people who don't use methamphetamine, as
> it's contrary to all evidence.
>
> If anyone can shed some light I'd be keen to hear your thoughts!
> Appreciate it :)
>
> Cheers,
> Jo
>
>
>
>
>
> --
> Haiyan
|