Hi Martyn,
Tom would have a better clue on this one,
but I would do:
EV1 EV2 EV3 EV4
1 0 1 -5.84
1 0 1 -6.84
1 0 -1 -20.84
1 0 1 -2.84
1 0 -1 -28.84
1 0 -1 7.16
1 0 -1 10.16
1 0 -1 18.16
1 0 1 13.16
1 0 -1 -0.84
1 0 1 9.16
0 1 1 -6.84
0 1 1 -0.84
0 1 1 -2.84
where EV3 is your gender effect that you need to demean (was too lazy to do so, sorry!)...
The contrasts are fine. Just don't forget to run randomise with the -D option to demean your data as well...
Hope this helps,
Gwenaelle
--- En date de : Mar 9.9.08, Martyn Ezra <[log in to unmask]> a écrit :
> De: Martyn Ezra <[log in to unmask]>
> Objet: [FSL] Regressors of no interest Age and gender in TBSS
> À: [log in to unmask]
> Date: Mardi 9 Septembre 2008, 18h03
> Hi
>
> I was hope someone could help me out.
>
> I have run TBSS using FSL 4.1.0 and want to compare a
> control group against
> a subject group.
> However the groups are not well matched for gender and age
> and i was hoping
> to make these regressors of no interest.
>
> Having read previous post this seems to have been done
> before by creating a
> separate GLM model and using the -x option in randomise.
> But i believe this
> has now changed?
>
> I have created the design matrix below (sample), where EV1
> is group A, EV2
> group B, EV3 being Male, EV4 being female and EV5 the
> demeaned ages.
>
> EV1 EV2 EV3 EV4 EV5
> 1 0 1 0 -5.84
> 1 0 1 0 -6.84
> 1 0 0 1 -20.84
> 1 0 1 0 -2.84
> 1 0 0 1 -28.84
> 1 0 0 1 7.16
> 1 0 0 1 10.16
> 1 0 0 1 18.16
> 1 0 1 0 13.16
> 1 0 0 1 -0.84
> 1 0 1 0 9.16
> 0 1 1 0 -6.84
> 0 1 1 0 -0.84
> 0 1 1 0 -2.84
>
> The contrasts are:
>
> Group A Group B
> Male Female
> Demeaned Ages
> C1 Group A > Group B 1 -1
> 0 0
> 0
> C2 Group B > Group A -1 1
> 0 0
> 0
>
> Is this the correct way of going about it?
>
> I would be very grateful if anyone has any suggestions
>
> Many thanks
>
> Martyn
|