Closed jdebacker closed 10 months ago
@john-p-ryan I think you looked into this and we decided to go with using the differences between bins. Is that right? Can you explain the reasoning here for future reference?
This issue has been resolved with PR #22, which (1) uses theoretical derivatives when possible and (2) finds that finite differences are quite accurate in our context in the cases they are used.
@john-p-ryan points out that we currently calculate derivatives using differences between discrete bins. However, one could have a more accurate approximation of the derivatives when using splines to approximate the distribution. We should test this alternative method and go with it if it proves more robust.