Which methods used in the "ksdensity" function for the selection of optimal bandwidth

"ksdensity" function returns the bandwidth of the kernel smoothing window. The default bandwidth is th optimal for normal densities.

Which methods used in the ksdensity for the selection of optimal bandwidth in the default bandwidth. (ex. Normal scale rules, Cross-validation; Least squares cross-validation, Biased cross-validation, ...)

Thank you.

Answer 1

[the cyclist]

If you look inside the ksdensity, you will see that it calls mvksdensity, which in turn calls the internal function apply_censoring_get_bandwidth, which calculates the bandwidth using "Scott's rule".

[Young Jin Kang]

Thanks for your answer. I looked inside the ksdensity. There is that optimal bandwidth is "u = sig * (4/(3*N))^(1/5)". But I know that this equation is "Silverman's rule". I don't know whether the equation in MATLAB is "Scott's rule" or "Silverman'r rule".

[the cyclist]

When I stated that that is "Scott's rule", I was just reading off the comment in the code. That comment could be out of date, I suppose, if the formula changed but the comment did not.

[Young Jin Kang]

Thanks for your answer.