Dear Kewei,
> We have two groups: normals and patients. Each subject (from either
> normal group or patient group) has two PET scans under each of two
> conditions (4 scans per subject, two conditions: hunger (H) and
> satiation (S)).
>
> We have compared the change of H vs S between the two groups, i.e.,
> (Hn-Sn)-(Hp-Sp) (Hn: hunger condition for normal subjects, others
> defined similarily). This analysis could be done very easily using
> SPM.
>
> Now, we would like to do the same comparison accounting for resting
> metabolic rate (RMR) as a covariate. The problem is that each subject
> has only one RMR reading.
>
> My question:
> (1) Can I enter the same RMR value four times (for each subject)
> corresponding to the four scans?
> (2) If not, what is the best way to do this kind of analysis?
In terms of the main effect of RMR, you have already modelled it
implicitly in the block partition of the confounds. You need do
nothing else. This is also the case for RMS x group interactions.
In terms of RMS x Condition interactions you would have to model these
explicitly as covariates (in a group specific fashion). They would
look some thing like:
For subject n (assuming H H S S for each subject) n - normal, p - patient
[RMR1n RMR1n -RMR1n -RMR1n RMR2n RMR2n -RMR2n -RMR2n...., 0 0 0 .... 0]';
[0 0 0 ... 0, RMR1p RMR1p -RMR1p -RMR1p RMR2p RMR2p -RMR2p -RMR2p ....]';
The contrast [0 ...0, 1 -1] would in this case give the RMS x Condition x
group interaction (if you wanted it).
I hope this helps - Karl
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|