I am trying to find out if there are any specific
rules when setting up contrasts in SPM2 for an fMRI
analysis where different conditions have differing
number of samples.
Suppose I have a de-meaned fMRI data set with three
conditions A, B and Rest with unequal number of scans
say 3, 2 and 4 respectively.
If my design matrix were as shown below (orthogonal,
full rank, simple box car)
TaskA TaskB Rest
1 0 0
1 0 0
1 0 0
0 0 1
0 0 1
0 0 1
0 0 1
0 1 0
0 1 0
My questions are:
Do the regressors need to be normalized in some way
(de-meaned/other) considering the fact that the actual
dataset (Y) is already de-meaned ? My concern here
being the differing number of samples for the
conditions.
Also, is it important to have the sum of the contrast
weighted regressors equal to zero i.e sum(c1*X1 +
c2*X2 + c3*X3) = 0 ? In other words is it ok to have
t-statistic contrasts such as [1 1 -2], [1 0 -1] and
[0 1 -1] in the case of the above design matrix ?
I did run an analysis comparing both ways i.e
regressors with zero mean and non-zero mean regressors
and got results which were very close but am trying to
verify whether it is ok to do so.
....Raj
____________________________________________________
Yahoo! Sports
Rekindle the Rivalries. Sign up for Fantasy Football
http://football.fantasysports.yahoo.com
|