Cv.warping(haerdle) | R Documentation |
Cross-validation for WARPing density estimation
Cv.warping(x, delta, kernel, Mstart=1, Mend)
x |
data vector |
delta |
step in bandwidth |
kernel |
coded value of kernel (1 to 5) |
Mstart |
left value of bandwidth is Mstart * delta |
Mend |
right value of bandwidth is Mend * delta |
matrix with two columns. The first is Mstart:Mend, the second the CV value.
`Smoothing Techniques with Implementation in S', Wolfgang Haerdle, Springer, 1991
#test of CV.Warping -- p.109 -- quite slow cv.simul<-function(n,k,seed) { .Random.seed <<- seed simul <- matrix(0,k,2) for (i in 1:k) { data <- runif(n)<=0.6 data <- data*(rnorm(n)-1)+(rnorm(n)+2)*(1-data) cv <- Cv.warping(data,0.1,5,Mstart=1,Mend=20) simul[i,] <- cv[sum((cv[,2]==min(cv[,2]))*c(1:20)),] } simul } seed.simul <- c(29,29,24,52,54,1,12,25,37,25,23,0) simulation <- cv.simul(400,100,seed.simul) mean(simulation[,1]) sqrt(var(simulation[,1]))