Derived quantities from kernel density estimates

Derived quantities from kernel density estimates.

Usage

dkde(x, fhat)
 pkde(q, fhat)
 qkde(p, fhat)
 rkde(n, fhat, positive=FALSE)

Arguments

x,q: vector of quantiles
p: vector of probabilities
n: number of observations
positive: flag to compute KDE on the positive real line. Default is FALSE.
fhat: kernel density estimate, object of class kde

Value

For the 1-d kernel density estimate fhat, pkde computes the cumulative probability for the quantile q, qkde computes the quantile corresponding to the probability p.

For any kernel density estimate, dkde computes the density value at x (it is an alias for predict.kde), rkde computes a random sample of size n.

Details

pkde uses the trapezoidal rule for the numerical integration. rkde uses Silverman (1986)'s method to generate a random sample from a KDE.

References

Silverman, B. (1986) Density Estimation for Statistics and Data Analysis. Chapman & Hall/CRC. London.

Examples

set.seed(8192)
x <- rnorm.mixt(n=10000, mus=0, sigmas=1, props=1)
fhat <- kde(x=x)
p1 <- pkde(fhat=fhat, q=c(-1, 0, 0.5))
qkde(fhat=fhat, p=p1)    
#> [1] -1.000396e+00 -1.039613e-05  5.002413e-01
y <- rkde(fhat=fhat, n=100)

x <- rmvnorm.mixt(n=10000, mus=c(0,0), Sigmas=invvech(c(1,0.8,1)))
fhat <- kde(x=x)
y <- rkde(fhat=fhat, n=1000)
fhaty <- kde(x=y)
plot(fhat, col=1)
plot(fhaty, add=TRUE, col=2)