Univariate kernel density estimation for bounded and unbounded support

Discrete variables are convoluted with the uniform distribution (see, Nagler, 2017). If a variable should be treated as discrete, declare it as ordered().

kde1d(x, mult = 1, xmin = -Inf, xmax = Inf, bw = NULL, bw_min = 0, ...)

Arguments

x: vector of length \(n\).
mult: numeric; the actual bandwidth used is \(bw*mult\).
xmin: lower bound for the support of the density.
xmax: upper bound for the support of the density.
bw: bandwidth parameter; has to be a positive number or NULL; the latter calls KernSmooth::dpik().
bw_min: minimum value for the bandwidth.
...: unused.

Value

An object of class kde1d.

Details

If xmin or xmax are finite, the density estimate will be 0 outside of \([xmin, xmax]\). Mirror-reflection is used to correct for boundary bias. Discrete variables are convoluted with the uniform distribution (see, Nagler, 2017).

References

Nagler, T. (2017). A generic approach to nonparametric function estimation with mixed data. arXiv:1704.07457

Examples

data(wdbc, package = "kdecopula")  # load data
fit <- kde1d(wdbc[, 5])            # estimate density
dkde1d(1000, fit)                  # evaluate density estimate
#> [1] 0.0003689227