It is probably not going to work very well if you have some classes that only have one member. You can eliminate members of classes that only have a few members. If you really need to have those cases, there are hybrid ways to analyze the data (but you can't use an FSGD at that point)
On 2/6/2022 10:41 PM, Zeng, Victor (BIDMC - Keshavan - Psychiatry) wrote:
External Email - Use Caution
Hi all,
I am doing an analysis and generating the FSGD. I understand sex and race are classes in this context, but something I was also hoping to regress out is the effects of scanner/sequences. We have a large participant size, so we have around 10+ scanner/sequences, which makes for a large amount of classes combinations, with some having just 1 (if not zero) fitting into a class.
Would you proceed with this nonetheless? If not, should scanner/sequence be something that we consider? Should we find some other way to adjust site (I don't think we can z-score...). Or can we consider site/scanner as a continuous variable
Victor Zeng Beth Israel Deaconess Medical Center Keshavan Lab --
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer