Dear SPM experts,
I need your advice about the following analysis with SPM99:
O15 PET scans of 6 subjects under two conditions: stimulator OFF and
ON. 2 repeat PET scans per condition. I want to identify correlation
between rCBF changes in ON-OFF images & ON-OFF clinical parameters.
Using Multi-subject, conditions & covariate model I entered both the
images and clinical scores as(Sub1OFFa Sub1OFFb Sub1ONa Sub1ONb
Sub2OFFa Sub2OFFb Sub2ONa Sub2ONb...total 24 scans + 24 scores).
Design description:
Global calculation: mean voxel value
Global mean scaling: (implicit in PropSca global normalization)
Global normalization: proportional scaling to 50
Parameters: 12 condition, 1 covariate, 6 block 0 nuisance
No interactions for the covariate and No covariate centering.
The design matrix:
2 columns indicating OFF/ON (conditions 0 0 1 1)
1 column of the covariate (24 values)
6 columns indicating the subject (6).
I used a contrast of (0 0 1) and (0 0 -1) to identify regions of
positive + negative correlation. The result was shown using a p
value of 0.01 (uncorrected). After examining the correlation for
the maxima of each cluster I was surprised that the results do
not always reach significance (i.e. p > 0.05). This was done
using the plot option of spm99. Would you please tell me whether
this design is appropriate and why I observe such a discrepancy
in individual peak points compared with SPM correlation maps?
Best regards and many thanks,
Masafumi Fukuda
*******************************************
Masafumi Fukuda MD
Functional Brain Imaging Laboratory
North Shore University Hospital
350 Community Drive
Manhasset NY 11030
TEL:516-562-1091
FAX:516-562-1008
e-mail:[log in to unmask]
********************************************
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|