Kernel support estimate

Kernel support estimate for 2 and 3-dimensional data.

Usage

ksupp(fhat, cont=95, abs.cont, convex.hull=TRUE)

# S3 method for class 'ksupp'
plot(x, display="plot3D", ...)

Arguments

fhat: object of class kde
cont: percentage for contour level curve. Default is 95.
abs.cont: absolute density estimate height for contour level curve
convex.hull: flag to compute convex hull of contour level curve. Default is TRUE.
x: object of class ksupp
display: one of "plot3D", "rgl" (required for 3-d only)
...: other graphics parameters

Value

A kernel support estimate is an object of class ksupp, i.e. a 2- or 3-column matrix which delimits the (convex hull of the) level set of the density estimate fhat.

Details

The kernel support estimate is the level set of the density estimate that exceeds the cont percent contour level. If this level set is a simply connected region, then this can suffice to be a conservative estimate of the density support. Otherwise, the convex hull of the level set is advised. For 2-d data, the convex hull is computed by chull; for 3-d data, it is computed by geometry::convhulln.

Examples

data(grevillea)
fhat <- kde(x=grevillea)
fhat.supp <- ksupp(fhat)
plot(fhat, display="filled.contour", cont=seq(10,90,by=10))
plot(fhat, cont=95, add=TRUE, col=1)
plot(fhat.supp, lty=2)


data(iris)
fhat <- kde(x=iris[,1:3])
fhat.supp <- ksupp(fhat)
plot(fhat)
plot(fhat.supp, add=TRUE, col=3, alpha=0.1)