Dear Steven,
>Could someone provide a simple explanation of the following design options
for a
>PET study. My study is an PET-FDG study with 2 scans per subject, with 1
>performance covariate collected per scan.
>
>1) Condition X subject interaction and covariate - What is the advantage
of this
>design ? I get an error message "unestimable condition effects"
>
The advantage of this design is that you allow for subject differences in
activation by estimating the activation effect separately for each subject.
Its main use is to create con*.img images to take to a second level
analysis, or to do conjunction across subjects.. In your case with only one
scan per subject you will only use up all your degrees of freedom and end
up with the error above.
>2) Centering options - what are the relative merits and conditions where
one
>would use overall centering of the co-variate, subject centering, and
condition
>centering. Which is the same as the procedure recomended by Andrew Holmes
for
>using difference covariate scores in SPM96 (mean center the difference
scores,
>divide by 2, alternately multiply by -1 and 1 for baseline and activation
>scans).
Generally speaking you should use overall centering when you have no
interaction with your covariate, subject centering when you have a
subject-by-covariate interaction and condition centering when you have a
condition-by-covariate interaction.
>
>Thanks.
>
>sg
>
Good luck Jesper
Jesper Andersson
Wellcome Dept. of Cognitive Neurology
12 Queen Square
London WC1N 3BG
phone: 44 171 833 7484
fax: 44 171 813 1420
|