Dear Didine,
> I am using SPM99 for a correlation study between binding of a
> radioligand and behavioral scores.
> I want to plot the data and compute the correlation coefficient using
> Matlab.
>
> I used VOI to extract values from a significant CLUSTER of voxels
> (instead of plot used for a single voxel).
>
> then
> if I DO NOT adjust data for contrast and type Y in the command window, I
> found r=0.77 from corrcoef(scores, Y)
> My understanding in this case is that data are only AnCova normalized
The function that implements this extraction is spm_region.m. If you
don't adjust for any F-contrast, the extracted data is based on the raw
data (i.e. the data you put into the general linear model) and the
global effects are not removed from the data.
>
> if I adjust data for contrast (effects of interest) and type Y in the
> command window then r=0.89 from corrcoef(scores, Y)
> My understanding is that data are AnCova normalized but in addition
> there is an adjustment for the contrast.
If you adjust for the default 'effects of interest' contrast, you remove
the global effects (i.e. any effects of no interest).
>
> What is the most appropriate result to report and why? can someone
> explain (math if possible) how the adjustment of data for contrast (eg
> effect of interest) is done and why or possibly provide a journal or
> book reference.
Check out the help for spm_regions. What I figure out from the
description is that the F-contrast specifies what effects you want to
keep, the others get removed from the data.
Note that if you adjust the data before you compute the correlation
coefficient, then you're talking about a correlation between your
covariate and the first component of adjusted data in a cluster. I think
it makes perfect sense to adjust the data before you compute the
correlation.
The other way would be to stick your covariate into the design matrix of
the general linear model and ask for the significance of the slope
parameter with a t-test or F-test and report a corrected p-value.
Stefan
--
Stefan Kiebel
Functional Imaging Laboratory
Wellcome Dept. of Cognitive Neurology
12 Queen Square
WC1N 3BG London, UK
Tel.: +44-(0)20-7833-7478
FAX : -7813-1420
email: [log in to unmask]
|