Hi,
The group design and contrasts in (1) look fine, although the group variable doesn't need to change (it is ignored by randomise).
In the correlation analyses I would recommend demeaning the values of your scores and adding back the EVs that model the mean (as in part 1). This then gives you the correlation centred about the mean of the score and the gray matter density (when you have a contrast that only includes the EV containing the demeaned scores). Otherwise your correlation analysis will largely just reflect the mean of the group (since that would be likely to be the strongest component of the original scores).
All the best,
Mark
On 28 Nov 2012, at 15:58, Jie-Yu Chuang <[log in to unmask]> wrote:
> ( Sorry, there is a problem with previous post, so I sent this new one )
> (^_^)
>
> [ Goals ]
> (1) do a VBM to compare 3 groups
> (2) do a correlational analysis ( scores in a psychological test versus MRI )
>
> Could you please take a look at my settings?
>
> [ FSL Glm Setting Up ]
> (1) 3 group VBM
> group EV1 EV2 EV3
> 1 1 0 0
> 1 1 0 0
> 2 0 1 0
> 2 0 1 0
> 3 0 0 1
> 3 0 0 1
>
> contrast EV1 EV2 EV3 Ftest
> 1 1 *
> 2 1 *
> 3 1 *
> 4 1 -1
> 5 1 -1
> 6 -1 1
> 7 -1 1
> 8 1 -1
> 9 -1 1
>
> (2) Correlational analysis --- all subjects as 1 group
>
> group EV1
> 1 56
> 1 89
> 1 33
> 1 55
> 1 33
> 1 83
>
>
> (3) Correlational analysis --- just focusing on group A
>
> group EV1
> 1 56
> 1 89
> 0 0
> 0 0
> 0 0
> 0 0
>
>
> Also, ALL subjects are included in the VBM template, but I only do a one or two group analysis.....will that be a problem?
> For example, I am analysing A group and B group, but the template contains information from C group as well???
>
> THANKS SO MUCH~~~~
>
|