Open jvaara opened 5 months ago
Given how the KDE is implemented in ArviZ, using a grid of points, we first need to consider how to account for the weights in the binning. This seems to be already solved in KDEpy https://github.com/tommyod/KDEpy/blob/master/KDEpy/binning.py
Certain Approximate Bayes Computation packages provide samples with weights; it would be useful to be able to use
plot_kde
directly for these applications.At least for the
stats.density_utils._fast_kde_2d
, it would be a straightforward modification.Cheers