Bandwidth selection for kernel conditional density estimation

 

We consider bandwidth selection for the kernel estimator of conditional density with one explanatory variable. Several bandwidth selection methods are derived ranging from fast rules-of-thumb which assume the underlying densities are known to relatively slow procedures which use the bootstrap. The methods are compared and a practical bandwidth selection strategy which combines the methods is proposed. The methods are compared using two simulation studies and a real data set.

Keywords: density estimation; kernel smoothing; conditioning; bandwidth selection.

R code

---

Errata:

• d should be |d| in the normal reference rule formulae.

Cite

@article{BH01,
  author    = {David M Bashtannyk and Rob J Hyndman},
  doi       = {10.1016/S0167-9473%2800%2900046-3},
  journal   = {Computational Statistics \& Data Analysis},
  number    = {3},
  pages     = {279--298},
  publisher = {Elsevier},
  title     = {Bandwidth selection for kernel conditional density estimation},
  volume    = {36},
  year      = 2001
}