Negative weights for KDE about kalepy

[lzkelley]

Hi, in principle it seems like this should be possible, but I've never explored negative weights. There will be some points in the code that need modification, for example, how does one calculate covariances using negative weights?

You're welcome to try implementing this in the code. If you disable the error message you're getting, it will allow negative weights, but then I think some numpy functions will probably throw errors.

[Pablobala]

Hi, just to let you know. I didn't find a way to handle negative weights because of the errors numpy throws when trying to calculate the covariances. So what I did is to split my sample into positive and negative weights, apply the KDE to both samples, and then rest the weights from "negative" sample renormalizing them by the sum of the sWeights. This seems to work but there are some downsides. If someone could find a way for handling negative weights it would be better.
Regards!