boot.lowess()

Sample the original data with replacement and fit LOWESS curves accordingly.

We keep on resampling the data and finally we will see several bootstrapped LOWESS curves, which may give us a rough idea about a confidence interval of the LOWESS fit.

library(animation)
ani.options(nmax = 100, interval = 0.02)

boot.lowess(cars, pch = 20, xlab = "speed", ylab = "dist")

plot of chunk demo-a

boot.lowess(cars, f = 1/3, pch = 20)

plot of chunk demo-b