Repository navigation
Covariance component models for REML available in nipy? #588
Description
Hello, thank you for this wonderful package!
I am trying to implement REML for covariance component calculation, and I confirm that nipy's reml works great, but I'm looking for an automated way of generating the components argument, similar to the functionality of spm_non_sphericity.m from SPM12. Is there similar functionality available in nipy?
The closest thing I've been able to find in nipy is the cov method of the RandomEffects class, but it doesn't yield the covariance components I expect for reml estimation of hyperparameters. I expect something similar to the attached screenshot, which is taken from this slideshow from www.tnu.ethz.ch
For simple models it is of course easy to manually specify the components, but I'm looking for something more automated, similar to spm_non_sphericity.m, for arbitrary factors / designs. Any advice you have would be most appreciated. Thank you!
Activity