Hi Emiliana -
there's certainly more than one approach to do this, but one simple
solution is to use the VOI button in SPM. Name the region, chose either
a sphere around an activation peak, or a whole cluster activated in your
regression model. In the Matlab workspace, you'll then find a structure
named xY. The [n x 1] vector xY.u is the first Eigenvariate, that gives
you a value for mean activation of the sphere/cluster in every single
subject. This vector can then be plotted against your behavioral
variable, within Matlab or by exporting it to any other statistical
program.
Hope this helps!
Maria
Emiliana Simon-Thomas schrieb:
> Hi -
> We've done a simple regression 2nd level analysis: 11 subjects con* images
> entered and a behavioral covariate entered into simple regression basic
> model. The SPM map is compelling, but we'd like to see a bivariate
> scatterplot of individual activation against the behavioral variable. My
> question is: how to get the individual activation values for such a plot?
>
> We tried getting t-scores from individual contrasts for voxels that survive
> a SVC - with the SVC volume being exactly the cluster of voxels that
> correlated with the behavioral measure in the group-wise basic model above.
> The resulting values (t-scores) do not appear to correlate with the
> behavioral variable. We tried this same SVC procedure with a couple other
> ROIs that overlap with the voxels that correlated with the behavioral
> variable in the group-wise basic model: again, the t-values don't seem to
> relate strongly with the behavioral measure.
>
> Is there another route to getting numbers that represent individual
> 'activation' in a given region for a given contrast that can be plotted
> against a behavioral variable of interest - and know that these values
> underlie the group-wise, simple regression finding?
>
> Thanks,
>
> Emiliana Simon-Thomas
> UC Berkeley
>
>
>
|