Dear Sean,
> Dear SPM community,
>
> My initial set of analysis using SPM99 as involved performing group
> comparisons on 20 controls, 48 Alzheimers and 23 Lewy Body subjects using
> the two sample t-test {one scan/subject}-design. Having
> carried this out, I would like to extend the analysis and include covariates
> in my design. The covariates I would like to include are Age (at scan time),
> MMSE scores and severity of disease at the
> time of scan.
>
> Can anybody suggest the best way to do this in SPM99?
>
There are a number of the PET models that will allow you to specify the design
you describe. I think the simplest way would be to use the "Single subject,
conditions and covariates" model (slightly counterintuitive, I admit).
There you would "pretend" all your data come from a single subject and enter
all your scans. You would then see group as a "condition" and enter a vector
like [1 1 ... 1 2 2 ... 2 3 3 ... 3] where the numbers indicate which of the
three groups the scan belongs to.
You will then be prompted for covariates, which you enter.
You will also be asked if you want a condition-by-covariate interaction, and if
you have a reason to believe that e.g. the dependence of blood flow on age is
different between the groups you will want that. Certainly you would assume
that the effect of disease severity is different between the two diseases, so
there you will want to include the interaction.
You will also be asked if you want to "center" your covariates around overall-
or condition-mean. Centering around condition means effectively orthogonalises
your covariates w.r.t. your conditions. I would suggest you do that.
>
> I shall look forward to hearing your very useful and appreciated comments!
>
I hope they were.
Good luck Jesper
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|