Dear Jennifer,
> We conducted a PET study to look at the effects of drug prior to and
> following surgery. Each patient had 15 scans before surgery and 15 scans
> after surgery. The first 5 scans are baseline, that is, no drug was
> administered ("off" scans). Following, the first 5 scans, the drug was
> injected and we have a variable number of "on" scans for each patient. To
> look at the differences between the "on" and "off scans the following was
> done:
>
> In SPM95, we entered the following commands:
>
> Statsitics
> select design type? Multisubject with replications
> # of subjects? 7
> # of conditions? 2
> selected "on" scans for condition 1 and "off" scans for condition
> 2 for each subject
> select global normalization? proportional scaling
> image left = subject left
> gray matter threshold? 0.8
> grand mean? 50
> # contrasts? 1
> [2]-contrast 1: 1 -1
>
> Is this correct for our contrast of interest? How does SPM95 deal with the
> variable number of scans for each subject? Does it take the average of the
> on scans and the average of the off scans for each subject?
There is no problem with the varying number of off-on-drug-scans, as
long as you specify these correctly in the design matrix. SPM (any
version I can remember) uses a general linear model to fit the data,
i.e. it estimates some parameters, which are weights of your basis
functions at each voxel. In other words, while taking into account any
other effect, e.g. subject effects, yes, SPM does something like finding
the average of an underlying drug-effect.
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]
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|