search

rs-station / careless

Merge X-ray diffraction data with Wilson's priors, variational inference, and metadata
MIT License
16 stars 6 forks source link

spearman CCpreds "look much different" from weighted pearson CCpreds #153

Closed hkwang closed 7 months ago

hkwang commented 8 months ago

With careless==0.4.1, I compared the careless.ccpred of two datasets with both-m spearman and -m pearson. These two datasts are Vukica's PYP dataset, processed with precognition for the careless paper (according to Doeke), and an e35_PDZ dataset processed with laue-dials.

CCpreds for PYP, weighted Pearson: image

CCpreds for PYP, Spearman: image

CCpreds for e35_PDZ, weighted Pearson: image CCpreds for e35_PDZ, Spearman: image

For each dataset, the CCpreds look "much different" between the two datasets.

kmdalton commented 8 months ago

@hkwang, is this at all unexpected? they're different estimators. do you think there is a bug?

kmdalton commented 7 months ago

I'm going to close this issue, because I don't see how it is a bug. Feel free to re-open if I missed something.