rmse_cvs: Estimate smoothing RMSE using leave-one-out cross-validation.
In hadley/bigvis: Tools for visualisation of big data sets

Description Usage Arguments See Also Examples

rmse_cv computes the leave-one-out RMSE for a single vector of bandwidths, rmse_cvs computes for a multiple vectors of bandwidths, stored as a data frame.

1
2
3

rmse_cvs(x, hs = h_grid(x), ...)

rmse_cv(x, h, var = summary_vars(x)[1], ...)

`x`	condensed summary table
`...`	other variables passed on to `smooth`
`h,hs`	for `rmse_cv`, a vector of bandwidths; for `rmse_cv` a data frame of bandwidths, as generated by `h_grid`.
`var`	variable to smooth

Other bandwidth estimation functions: best_h; h_grid

set.seed(1014)
# 1d -----------------------------
x <- rchallenge(1e4)
xsum <- condense(bin(x, 1 / 10))
cvs <- rmse_cvs(xsum)

if (require("ggplot2")) {
autoplot(xsum)
qplot(x, err, data = cvs, geom = "line")
xsmu <- smooth(xsum, 1.3)
autoplot(xsmu)
autoplot(peel(xsmu))
}

# 2d -----------------------------
y <- runif(1e4)
xysum <- condense(bin(x, 1 / 10), bin(y, 1 / 100))
cvs <- rmse_cvs(xysum, h_grid(xysum, 10))
if (require("ggplot2")) {
qplot(x, y, data = cvs, size = err)
}