Hi there
> I have a question in regard to our study design for VBM using SPM2.
> We have 2 groups (controls and patients) and have measured a behavioral
> score (continuous variable) for all subjects.
> We wish to know the following:
> 1) Among all subjects combined, where Is a correlation present between
> the behavioral score and gray matter volume? For this, we combine all
> subjects and use a simple regression model.
> 2) Where does the correlation between gray matter volume and behavioral
> score differ between the two groups?
>
> The 2nd Q is the main question. Our best guess as to how to address Q2 is
> to use an ANCOVA, with two groups and using the behavioral score as nuisance
> variable. In this scenario, we use the contrast [0 0 0 1].We expect that
> this would treat the behavioral score as the main variable of interest and
> the groups as nuisance variables.
> Is this the right study model and contrast?
>
I would go for a multiple regression model (no cste)
what you put in is the group (1111-1-1-1-1), then the scores (xxxxx0000
and 00000xxxx) and of course the nuisance (ie global volume, age, etc) -
by doing that you can look at differences between groups (1),
correlation in each group (01 and 001) and correlation across groups
(011) and differences between groups (01-1)
hope this answers your question
cyril
|