ageller
commented
2 years ago I created another option for fitting PMs in 2 steps (field first and then cluster). That also requires lots of hand holding and doesn't always work.
I wonder if we should make another option to just fit single 2D Gaussian to the cluster members, as identified by RV and/or distance, in PM space and then take some region around that or use the Gaussian to produce some membership probability (e.g., normalize the Gaussian to a max of 1 at the center and then just take the value at that PM as the membership). I have a code that does this in my NGC6819test directory (but not yet on GitHub).
At least for NGC 188, the fit for PM was very dependent on the initial guess for the mean and the fitting range. I don't want that much hand holding for all clusters, so we need a better fitting method. That same method should work for RV (which for NGC 188 did not require any hand holding). I think the distance guess can probably come from my OC summary table, assuming it is accurate enough.