I rediscovered today that I cannot assess group differences within
condition and condition effects at the same time. I am hoping to design a
way to do so with SPECT data.
I am looking for a treatment effect and to assess the difference between
treated/untreated groups.
In addition, the hypothesis of treatment effect relates to the degree of
abnormality at baseline.
to compare the effect of treatment over time in patients to a control
group, I set up a ranFx analysis of
[-1 1 1 -1] or the inverse for decreases.
however, I am interested in how the toography of these effects, *for each
subject*, relate to abnormalities at the time of the first scan--
[[-1 0 1 0] which cannot be estimated as such, a failed effort at fitting a
parameter across partitions of the design matrix.
So, If I separately compare groups at baseline, I can generate a statistic
image of baseline decreases relative to controls--
[-1 1] my contrast of interest
How do I constrain the analysis of changes with treatment to these areas?
In general , I believe I cannot mask one contrast with a contrast from a
separate analysis.
If I were to try to use each subject's baseline abnormality image as a mask
for the change analysis itself, I am concerned that I will have limited my
search volume so much that it will artificially inflate my corrected voxel
statistics.
Is there a workable solution to this?
********************************************************
Christopher Gottschalk, MD *
Assistant Professor of Neurology & Psychiatry *
Yale School of Medicine *
*
Mailing Adress: *
VAMC [116-A] tel [203] 932-5711 x4329*
950 Campbell Avenue FAX 937-4791*
West Haven, CT 06516 *
********************************************************
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|