GUI Density Estimation using Tcl/Tk

This is graphical user interface (GUI) to density, allowing for dynamic bandwidth choice and a simple kind of zooming, relying on library(tcltk).

Usage

tkdensity(y, n = 1024, log.bw = TRUE, showvalue = TRUE,
          xlim = NULL, do.rug = size < 1000, kernels = NULL,
          from.f = if (log.bw) -2 else 1/1000,
          to.f   = if (log.bw) +2.2 else 2,
          col = 2)

Arguments

y

numeric; the data the density of which we want.

n

integer; the number of abscissa values for density evaluation (and plotting).

log.bw

logical; if true (default), the gui scrollbar is on a log bandwidth scale, otherwise, simple interval.

showvalue

logical; if true, the value of the current (log) bandwidth is shown on top of the scrollbar.

xlim

initial xlim for plotting, see plot.default.

do.rug

logical indicating if rug(y) should be added to each plot. This is too slow for really large sample sizes.

kernels

character vector of kernel names as allowable for the kernels argument of the standard density function.

from.f, to.f

numeric giving the left and right limit of the bandwidth scrollbar.

col

color to be used for the density curve.

Details

library(tcltk) must be working, i.e., Tcl/Tk must have been installed on your platform, and must have been visible during R's configuration and/or installation.

You can not only choose the bandwidth (the most important parameter), but also the kernel, and you can zoom in and out (in x-range only).

Value

none.
(How could this be done? tcltk widgets run as separate processes!)

Author

Martin Maechler, building on demo(tkdensity).

Examples

if (dev.interactive(TRUE)) ## does really not make sense otherwise
 if(try(require("tcltk"))) { ## sometimes (rarely) there, but broken

  data(faithful)
  tkdensity(faithful $ eruptions)

  set.seed(7)
  if(require("nor1mix"))
     tkdensity(rnorMix(1000, MW.nm9), kernels = c("gaussian", "epanechnikov"))
 }