On Wed, Mar 18, 2009 at 10:22 AM, Greig de Zubicaray <[log in to unmask]> wrote:

Hello list,

I've just run a multiple regression with tbss on some skeletonised FA data in a single group of subjects using 3 EVs - age, 1 target behavioural variable and 1 potentially confounding behavioural measure. I created a design matrix with the 3 normalised EVs, and orthogonalised the target EV with respect to age and the confounding EV. I assumed this would show me the changes in FA attributable to the target EV with the other 2 covaried out. This was using randomise and the TFCE option. When I display the tbss_tfce_corrp_tstat contrast image in fslview with the -b 0.95,1 threshold, most of the skeletonised data appears to correlate with the target variable. This is the case even when I use a more conservative 0.99,1 threshold.

I found this a little surprising, so ran the same design again without orthogonalising the target EV. The results were virtually the same.

So I guess the question is, have I done anything wrong here?

thanks in advance,

Greig

--
Dr Greig de Zubicaray | fMRI Laboratory
Centre for Magnetic Resonance | University of Queensland | Brisbane, QLD 4072 | AUSTRALIA
Phone: +61-7-3365-4250 (direct) | +61-7-3365-4100 (reception) | Fax: +61-7-3365-3833
Email: [log in to unmask] | Web: http://www.fmrilab.net/
CRICOS Provider Number 00025B

--
____________________________________________
Thomas Nichols, PhD
Director, Modelling & Genetics
GlaxoSmithKline Clinical Imaging Centre

Senior Research Fellow
Oxford University FMRIB Centre