For univariate kernel density estimators (KDE), I use Silverman's rule for calculating $h$:

\begin{equation} 0.9 \min(sd, IQR/1.34)\times n^{-0.2} \end{equation}

What are the standard rules for multivariate KDE (assuming a Normal kernel).


For a univariate KDE, you are better off using something other than Silverman's rule which is based on a normal approximation. One excellent approach is the Sheather-Jones method, easily implemented in R; for example,

plot(density(precip, bw="SJ"))

The situation for multivariate KDE is not so well studied, and the tools are not so mature. Rather than a bandwidth, you need a bandwidth matrix. To simplify the problem, most people assume a diagonal matrix, although this may not lead to the best results. The ks package in R provides some very useful tools including allowing a full (not necessarily diagonal) bandwidth matrix.


For univariate kernel density estimation, the bandwidth can be estimated by Normal reference rule or Cross Validation method or plug-in approach.

For multivariate kernel density estimation, a Bayesian bandwidth selection method may be utilized, see Zhang, X., M.L. King and R.J. Hyndman (2006), A Bayesian approach to bandwidth selection for multivariate kernel density estimation, Computational Statistics and Data Analysis, 50, 3009-3031


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.