A penalised-spline alternative to the kernel methods above. Fits
y ~ s(x, k = k) and returns fitted values at x_eval.
Usage
gam_smoother(x, y, x_eval = NULL, k = 10, family = stats::gaussian())
Arguments
- x
Numeric covariate vector.
- y
Numeric outcome vector.
- x_eval
Evaluation grid (defaults to x).
- k
Basis dimension for the smoother (default 10).
- family
GLM family for mgcv::gam (default
gaussian()).
Value
A list with fit (the fitted gam object),
x_eval, y_hat (predictions), and edf
(effective degrees of freedom).
Examples
if (requireNamespace("mgcv", quietly = TRUE)) {
set.seed(1)
x <- sort(runif(40, -2, 2))
y <- sin(x) + 0.2 * rnorm(40)
xe <- seq(-1.5, 1.5, length.out = 20)
r <- gam_smoother(x, y, x_eval = xe, k = 5)
head(r$y_hat)
}
#> [1] -0.9758711 -0.9677393 -0.9382219 -0.8824320 -0.7990992 -0.6891929
