Dear colleagues,
We have acquired an FDG-PET data-set in 8 subjects under 4 conditions: vehicle,
and three different doses of a drug. I have previously used SPM to obtain
t-test contrasts after running a simple multisubject/multicondition design but
I have not yet examined the effect of dose in a parametric manner. I would like
to try the following:
Use dose as a covariate and explore linear, log and parabolic relationships. How
do I ascertain what function may be the best fit and how do I interpret the
image results accordingly? For example, if I use a parabolic function where the
medium dose is ascribed the highest covariate value, would I expect to see
voxels/clusters which showed this relationship?
Use plasma exposure (PK) as a correlate against all images. Again, would I have
to run the values as a function of PK in order to explore log, linear, and
parabolic relationships?
Am I able to run contrasts post-hoc to see negative relationships or do I have
to re-run everything again using negative values?
Thanks for any help you can give and any useful references you might know of.
Cheers, Graham
|