Dear SPMers,
We're using SPM99 in different PET studies. After we finish the statistical
model estimation, we try to use Matthew's worleyize program to pool the
variance. For some study, this program did reduce the noise and improve the
result. For some study, it just changes the result a little bit. The program
pools the variance through averaging the Residuals sum-of-squares for all
analyzed voxels. So it pooled the variance across brain of all subjects. The
assumption is that the variance for all subjects' brain remain the same. Is
this assumption too strict? If we use the assumption that the variance for
each subject's brain remain the same, can we use this worsleyize program?
Thanks a lot.
Sincerely,
Yanjun
|