Hello,
I've put some effort into deciphering spm_est_nonsphericity and spm_reml recently, but there are still some things I don't understand. Can someone help me out?
– spm_est_nonsphericity selects voxels for AR estimation based on an omnibus F-contrast. What's the rationale behind this? Wouldn't we want to estimate the error nonsphericity on average across all voxels that enter subsequent analysis?
– spm_reml estimates covariance components based on Cy, the full data temporal covariance (averaged across selected voxels). Isn't it a problem that the data covariance contains not only the error covariance, but also the effect covariance? Does spm_reml account for this? If not, why not? If yes, how?
Thanks a lot!
Carsten
--
Carsten Allefeld, M.A., Dr. rer. nat.
Research Associate
Bernstein Center for Computational Neuroscience
Theory and Analysis of Large-Scale Brain Signals
Charité – Universitätsmedizin Berlin
Philippstr. 13, Haus 6
10115 Berlin, Germany
http://www.carsten-allefeld.de/
|