kde(haerdle) | R Documentation |
kernel density estimation
kde(data, kernel, h, points=100)
data |
vector of data |
kernel |
coded kernel: = Uniform, 2 = Triangle, 3 = Epanechnikov, 4 = Quartic, 5 = Triweight, 6 = Gaussian, 7 = Cosinus |
h |
bandwidth |
points |
number of points at which to evaluate the density. |
matrix with two columns giving the gridpoint and density estimate there.
Evaluates on a grid covering the range of the data.
`Smoothing Techniques with Implementation in S', Wolfgang Haerdle, Springer, 1991
data(buffalo) # Figure 2.8 histogram(buffalo,8,0, ylim=c(0,0.022)) lines(kde(buffalo,3,8,100)) data(faithful) # Q2.2 (see pp.215-6) plot(kde(faithful$eruptions, 4, h=0.1, 80), type="l") plot(kde(faithful$eruptions, 4, h=0.3, 80), type="l") plot(kde(faithful$eruptions, 4, h=0.6, 80), type="l") data(dat.mixed) # Figure 2.11 test.kde<-kde(dat.mixed,6,0.493,100) grid<-test.kde[,1] true.density<- (0.6*exp(-0.5*(grid+1)^2) + 0.4*exp(-0.5*(grid-2)^2))/sqrt(2.0*pi) density.mixed<-matrix(c(grid,true.density), length(grid), 2) plot(density.mixed,type="n",xlab="dat.mixed",ylab="",cex=0.6) lines(test.kde) lines(density.mixed, lty=3)