account for multiplicity in spot prediction/res-dep BW

rs-station / laue-dials

A package for analyzing Laue x-ray crystallography data using the DIALS framework.

https://rs-station.github.io/laue-dials/

MIT License

5 stars 3 forks source link

account for multiplicity in spot prediction/res-dep BW #54

Open DHekstra opened 9 months ago

DHekstra commented 9 months ago

since we have miller indices and spacegroup by the time we predict spots, we know the multiplicities of each spot. perhaps there is merit in adjusting the threshold accordingly. (granting that will require some thought to determine what "accordingly" means)

kmdalton commented 9 months ago

This should be straightforward to implement. I'm not sure if it'll make much difference, but it's an interesting idea.

DHekstra commented 9 months ago

would it make sense to replace the KDE with logistic regression? with a set of strong reflections and a set of not strong reflections as the training data and wavelength, rlp, multiplicity, and possibly the image number as predictive variables?