Dear SPMers
Currently I'm working on ICA of SPECT brain images.
I has split the 20 brain images into 4 quartile groups. The calculated the
independent components for each group then I got 5 independent components for
each group. The independent components are then compared across groups using
SPM. Then I changes independent components into maps. To compare the
components in SPM,
Q1: Can I put the component maps into the formula directly for two-sample
t-test?
(i.e put 5 independent component maps into Group1, and put another 5
independent component maps into group2)
Q2: If I can, Then I want to find the activation region when I comparing
groups. So I did t-contrast with [1 -1] for Group1> Group2, or [-1, 1] for
Group1< Group2. Am I right?
Q3: If I want to " multiple regression", Can I put in all 20 independent
component maps(i.e 5 independent component maps for Group1, 5 independent
component maps for Group2 etc)?
then, what contrast vector I should use to do t-contrast ?
(i.e. I used [1 -1 0 0] for Q1<Q2, and [0 -1 1 0 0] for Q2<Q3 etc)
Anyone point my mistakes out please and
any advice will be appreciate.
Cheers
==========================================
In Kang
Ph.D Statistics Student
Department of Mathematics and statistics
University of Canterbury
Private Bag 4800
Christchurch New Zealand
Tel: +64 3 364 2987 ext 8873
|