Hello all,
In a recent grant submission, I was asked to provide efficiency estimates
for my parametric modulators. In the proposed studies, I have 4 parametric
regressors (modulators), each with 5 levels.
I was able to calculate from the equations provided on the listserve:
% define your contrast of interest
contrast = ;
load SPM
Var = contrast'*SPM.xX.Bcov*contrast;
E = 1/Var
But, these estimates seem to be influenced heavily by the scaling of the
continuous variables (if I scale 0 to 100, or 0 to 1, I get different results).
I have now run two of the studies, and both have significant and replicated
results at the 2nd level using the estimates from the parametric modulators
as input (they are orthogonal to one another by design), which suggests to
me that there is plenty of power.
My question here is: Is there a way to calculate the efficiency for a
parametric modulator in a way that provides non-arbitrary and meaningful
estimates for a grant resubmission? I fear that I may be missing something
critical here.
Thanks for any advice that you have.
Best,
Wil
|