Confidence/prediction intervals for predictions

[slagtermaarten]

Hi, thanks for developing this package first and foremost!

I'd like to have confidence intervals based on the weighted means of the k nearest neighbors for my continuous response variables - there are virtually no replicates in my training points and bootstrapped CIs do not seem to be an option for me. I've just augmented the predict.kknn method and kknn function to get a table of CIs for continuous variables but the method I'm using is probably not optimal yet. Would you be interested at all in merging something like this into your master branch? If so, I'd love to work it out more fully. My fork is here