predict.smooth.spline {stats}R Documentation

Predict from Smoothing Spline Fit


Predict a smoothing spline fit at new points, return the derivative if desired. The predicted fit is linear beyond the original data.


## S3 method for class 'smooth.spline':
predict(object, x, deriv = 0, ...)


object a fit from smooth.spline.
x the new values of x.
deriv integer; the order of the derivative required.
... further arguments passed to or from other methods.


A list with components

x The input x.
y The fitted values or derivatives at x.

See Also



cars.spl <- smooth.spline(speed, dist, df=6.4)

## "Proof" that the derivatives are okay, by comparing with approximation
diff.quot <- function(x,y) {
  ## Difference quotient (central differences where available)
  n <- length(x); i1 <- 1:2; i2 <- (n-1):n
  c(diff(y[i1]) / diff(x[i1]), (y[-i1] - y[-i2]) / (x[-i1] - x[-i2]),
    diff(y[i2]) / diff(x[i2]))

xx <- unique(sort(c(seq(0,30, by = .2), kn <- unique(speed)))) <- match(kn, xx)# indices of knots within xx
op <- par(mfrow = c(2,2))
plot(speed, dist, xlim = range(xx), main = "Smooth.spline & derivatives")
lines(pp <- predict(cars.spl, xx), col = "red")
points(kn, pp$y[], pch = 3, col="dark red")
mtext("s(x)", col = "red")
for(d in 1:3){
  n <- length(pp$x)
  plot(pp$x, diff.quot(pp$x,pp$y), type = 'l', xlab="x", ylab="",
       col = "blue", col.main = "red",
       main= paste("s",paste(rep("'",d), collapse=""),"(x)", sep=""))
  mtext("Difference quotient approx.(last)", col = "blue")
  lines(pp <- predict(cars.spl, xx, deriv = d), col = "red")

  points(kn, pp$y[], pch = 3, col="dark red")
  abline(h=0, lty = 3, col = "gray")
detach(); par(op)

[Package stats version 2.5.0 Index]