Calculation of the ICV bandwidth for the Gaussian density estimator corresponding to expression (12) of Savchuk, Hart, and Sheather (2010).
Usage
h_ICV(x)
Arguments
x
numerical vector of data.
Value
The ICV bandwidth.
Details
Computing the ICV bandwidth for a univariate numerical data set of size $n<12,058$. the="" icv="" bandwidth="" is="" consistent="" for="" mise="" optimal="" (see="" wand="" and="" jones="" (1995)).="" gaussian="" kernel="" used="" computing="" ultimate="" density="" estimate.="" following="" values="" of="" paramaters="" selection="" L_ICV are used: $(\alpha,\sigma)=(2.42, 5.06)$. The ICV bandwidth does not exceed the oversmoothed bandwidth of Terrell (1990). See expression (12) of Savchuk et al. (2010).
References
Savchuk, O.Y., Hart, J.D., Sheather, S.J. (2010). Indirect cross-validation for density estimation. Journal of the American Statistical Association, 105(489), 415-423.
Wand, M.P. and Jones, M.C. (1995). Kernel Smoothing. Chapman and Hall, London.
Terrel, G. (1990). The maximum smoothing principle in density estimation. Journal of the American Statistical Association, 85, 470-477.