## ----------------------------------------------------------------------------
##   R scripts for Haerdle/Mueller/Sperlich/Werwatz: "Nonparametric and
##   Semiparametric Modelling", Springer Series in Statistics, 2004
## ----------------------------------------------------------------------------
##   Script      SPMkdeconstruct-Sliders
## ----------------------------------------------------------------------------
##   Description visualizes the construction of a kernel density
##               estimate (with additional slider to choose bandwidth
##               and sample size)
## ----------------------------------------------------------------------------
##   Author      Marlene Mueller, 2010/07/11
## ----------------------------------------------------------------------------

library(tcltk)
library(TeachingDemos)

kde.refresh <- function(...) {

  n <- slider(no=2)
  set.seed(19240215)
  x <- rnorm(n)
  h <- slider(no=1)

  xr <- diff(range(x))
  ng <- 100
  xg <- (xr+7*h)*(0:(ng-1))/(ng-1) + min(x)-3.5*h

  fk <- matrix(0,nrow=ng,ncol=n)
  for (j in 1:n){
    fk[,j] <- dnorm((xg-x[j])/h)/(n*h)
  }
  fh <- rowSums(fk)

  ylim <- c(-0.1*max(fh),max(fh))

  plot(xg,fh,type="l",lwd=2,ylim=ylim,
       main=paste("Construction of Density Estimate (n=",n,")",sep=""),
       xlab="Data x",ylab="KDE fh, Weights Kh")
  points(cbind(x,rep(-0.05*max(fh),n)),col="red",pch=3)

  for (j in 1:n){
    lines(xg,fk[,j],col="blue")
  }
}

slider(sl.functions=kde.refresh,
       sl.names=c("Bandwidth", "Sample Size"),
       sl.mins=c(0.01, 1),
       sl.maxs=c(2, 100),
       sl.deltas=c(0.01, 1),
       sl.defaults=c(0.25, 10),
       title = "Choose KDE Construction Parameters")