density: Density class

In BillVenables/WWRGraphics: Traditional Graphics Enhancements for WWR

Description

Promotion of an S3 class to virtual S4

Usage

density(x, ...)

## S4 method for signature 'numeric'
density(
  x,
  n = 4096,
  from,
  to,
  ...,
  lower = -Inf,
  upper = Inf,
  normalize = TRUE,
  eps = .Machine$double.eps
)

## S4 method for signature 'formula'
density(x, data = parent.frame(2), ...)

Arguments

x	numeric object or formula
...	additional arguments to be passed on to methods
n	integer: number of points to use in the output
from, to	lower and upper values of x to use in the output
lower, upper	lower and upper known limits to the support, and hence points at which to fold the density
eps	just the tiniest bit, so tiny you would never see it in a million years...
data	data frame or environment from which to get the data
normalise	logical: should the result be re-normalized to have total area unity?

Value

a density object

Examples

set.seed(123)
oldPar <- par(bg = "beige", font = 2, mar = c(5,5,4,1)+0.1, las = 1)
for(N in c(50, 500, 5000)) {
  x <- rexp(N, 2)
  dx <- density(x)
  dx0 <- update(dx, lower = 0)
  plot(dx0, col = "red", lwd = 2, bty = "n", main = "KDE Estimation",
       xlab = "x", sub = paste("Sample size =", N),
       ylab = expression(f[X](x)*',  '*X*' ~ '*exponential(2)),
       xlim = xrange(dx, dx0),
       ylim = yrange(dexp(0, 2), dx, dx0))
  lines(dx, col = "grey", lwd = 2)
  curve(dexp(x, 2), add = TRUE, n = 1000, col = "dark green", xpd = NA)
  lines(density_histogram(x, binWidth(0.125)), col = alpha("blue", 0.5))
  rug(x, col = alpha("red", (order(order(x))/N)^0.25))
  grid()
  legend("center", text.font = 2,
         c("naive KDE", "KDE folded at 0", "naive density histogram", "true density"),
         col = c("grey", "red", alpha("blue", 0.5), "dark green"),
         lty = "solid", lwd = 2, cex = 0.8, bty = "n",
         title = "Method")
}
par(oldPar)

BillVenables/WWRGraphics documentation built on Feb. 1, 2021, 12:13 p.m.