kds.ci(haerdle) | R Documentation |
Confidence intervals for kernel density estimates
kds.ci(data, h, g, x, alpha)
data |
data vector |
h |
bandwidth for estimator |
g |
bandwidth for second derivative estimator |
x |
vector of x-values for confidence intervals |
alpha |
1 - confidence coefficient |
matrix with columns x, the two estimators, lower ci, upper ci, one row for each point in x.
`Smoothing Techniques with Implementation in S', Wolfgang Haerdle, Springer, 1991
data(dat.mixed) # Figure 2.12 (p.63-4) grid <- c(0:10)/2-2 ci <- kds.ci(dat.mixed,0.493,0.5,grid,0.05) yli <- c(0,max(ci[,5])) plot(ci[,1], ci[,2], type="l",cex=0.6, ylim=yli) for(i in 1:11) # not 20 { lines(rep(ci[i,1],2), c(ci[i,4],ci[i,5])) lines(ci[i,1]+c(-0.03,0.03), rep(ci[i,4],2)) lines(ci[i,1]+c(-0.03,0.03), rep(ci[i,5],2)) } grid <- c(0:100)/20-2 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) lines(density.mixed, lty=3)