Hi Suman,
not quite: if you expect an interaction of gender x group, then you should indeed split the gender covariate as you did, but then *separately* demean EV3 (so gender in group 1 *only*) and EV4 (gender in group 2 *only*).
On the other hand, if you don't expect any interaction, then gender should be in one column and demeaned across the two groups...
Cheers,
Gwenaelle
> De: Suman Sen <[log in to unmask]>
> Objet: Re: [FSL] Re : [FSL] Help - VBM Randomise Matrix
> À: [log in to unmask]
> Date: Vendredi 11 février 2011, 0h45
> Thanks Gwenaëlle.
> My problem however is that in my group there is unequal
> number of males and females and the total brain volume is
> significantly different for males than females. I have read
> in multiple places that sex can be a confounder in VBM
> analysis unless it is corrected. What should be the proper
> approach for correcting this? On approach I can think of is
> to just add a nuisance variable for sex and use it as a
> covariate? In that case the matrix will be something like
> this:
>
> Groups
> EV1 EV2 EV3 EV4
> 1 1
> 0 0.48 0
> 1 1
> 0 -0.53 0
> 1 1
> 0 -0.53 0
> 2 0
> 1 0
> 0.53
> 2 0
> 1 0
> -0.48
> 2 0
> 1 0
> 0.53
>
>
> Where,
> EV1 = mean of group 1,
> EV2 = mean of group 2,
> EV3 = sex (demeaned) in group 1,
> EV4 = sex (demeaned) in group 2.
>
> The final contrasts are
>
> EV1
> EV2 EV3
> EV4
> 1>2 1
> -1
> 0 0
> 2>1 -1
> 1
> 0 0
> 1>2 adj 0
> 0
> 1
> -1
> 2>1 adj 0
> 0
> -1 1
>
> Does this look alright? Do I have to specify contrasts for
> EV3 and EV4 to account for variability in sex or are the
> first two contrasts alone enough?
>
> Thanks
> Suman
--------------------------------------------------------------------
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
--------------------------------------------------------------------
|