LogConcDEAD (version 1.1-1)

plot.LCDweights: Plot function for weighted log-concave estimator

Description

This plot function currently only plots 1- and 2-d data.

Usage

## S3 method for class 'LCDweights':
plot(x, uselog=FALSE, method="akima", itype="p",
addp=TRUE, gridlen=100, g=NULL, marg=NULL, g.marg=NULL, ...)

Arguments

x

Object of class LCDweights (e.g. as output from lcd.weights)

uselog

logical indicating whether the log or true scale should be plotted

method

The akima package is used for linear interpolation

itype

Plot type: p perspectice, c contour, i image, ic image and contour, r using rgl (the best!)

addp

logical: should the data points be plotted on the surface?

gridlen

Number of points at which the MLE is evaluated in each direction

g

If many plots are required, it may be quicker to compute the grid using lcd.interp(out) and pass the result to plot

marg

axis-aligned marginals can be plotted (set to 1 or 2)

g.marg

may be quicker to compute marginal using lcd.marg or lcd.marg2 and pass to plot

...

Other plot arguments

Details

This function makes use of the akima package.

For one-dimensional data, the log estimator is piecewise linear with knots at observation points.

The same is (roughly speaking) true in higher dimensions. The estimator is evaluated on a grid using the interp function from akima.

To come: plotting of marginals for higher-dimensional data.

See Also

interp

Examples

## some simple normal data

set.seed(101)
x <- matrix(rnorm(200), ncol=2)
w <- runif(100)
w <- w/sum(w)
out <- lcd.weights(x,w)
par(mfrow=c(2,2))
plot(out,  itype="c", main="Density")
plot(out,  itype="c", uselog=TRUE, main="LogDensity")
plot(out,  itype="i", main="Density")
plot(out,  itype="i", uselog=TRUE, main="LogDensity")