Dear SPMers
I have a question about 'AnCova' test.
I have studied about cognitive impairments of SLE patients
by using brain SPECT images, neuropsychological tests and SPM2.
As a preliminary study, I would like to compare SLE patients (all female, n=19)
vs. normal control (all female, n=10).
However, 'age' distribution is somewhat different between patients and
controls. So I think that '1scan/subject(AnCova)' is statistically appropirate
option, not '1scan/subject(two sample t-test)' option in this situation.
I want you to confirm whether my idea and design descriptions in SPM are
right or not. Help me, please.
I desinged as bellows...
(PET)
comapre-population: 1scan/subject(AnCova)
--> group 1 (10 control sujects)
--> group 2 (19 patients)
--> nuisance variable '1' and input'ages' of 29 subjects.
--> nuisance variable name 'age'
--> AnCova for global normalization
--> grand mean scaling : scaling of overall grand mean to '50'
--> threshold masking --> relative --> '0.8'
--> mean voxel value (within per image fullmean/8mask)
--> nonsphericity correction 'no'
Design description...
Design : Compare−populations: 1 scan/subject (AnCova)
Global calculation : mean voxel value (within per image fullmean/8 mask)
Grand mean scaling : scaling of overall grand mean to 50
Global normalisation : AnCova
Parameters : 2 condition, +0 covariate, +1 block, +2 nuisance
5 total, having 4 degrees of freedom
leaving 25 degrees of freedom from 29 images
Thanks in advance for your answers & suggestions,
Oh
|