Dear Mirko,
> I would like to know if there is any way to make an estimate of the
> sample size required to observe an effect between two groups in a
> cluster of 100 woxels with and p<0.001 using SPM. I was looking
> through course notes by I was not able to find any where reference to
> this problem.
No I am afraid not. To estimate sample sizes (or more generally
sensitivity and specificity for a given sample size) one needs to know
the effect size under the alternate hypothesis. We do not. If you are
interested in estimating the sensitivity (for a given n and modeled
effects of a distributed sort) I would read:
Friston KJ Holmes A Poline J-B Price CJ Frith CD. Detecting
activations in PET and fMRI: levels of inference and power.
NeuroImage, 1996;4:223-235
I hope this helps - Karl
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|