This function retrieves data for a specified smooth term from a fitted gam object (package mgcv) which is handy when you want to graph the smoother with ggplot.

example.R

library(ggplot2)
library(mgcv)

# Toy data set
dat <- data.frame(
  x = seq(0, 2*pi, length.out = 100),
  y = sin(3*x) / sin(x/2)
)

# Randomly thin the data
dat <- dat[sample(1:nrow(dat), nrow(dat)/3), ]

# Fit the model
m <- gam(y ~ s(x), data = dat)

# Plot the smooth term with plot.gam
plot(m, shade = TRUE, residuals = TRUE, cex = 5)

# Plot the smooth term with ggplot
smooth.dat <- getSmoother(m, 1)

ggplot(data = smooth.dat, aes(x = x)) +
  # smooth term bounds and line
  geom_ribbon(aes(ymin = lwr, ymax = upr), fill = "#ccccff") +
  geom_line(aes(y = fit), colour = "blue") +
  
  # data points
  geom_point(data = dat, aes(x = x, y = y)) +
  
  theme_bw()

getsmoother.R

library(mgcv)

getSmoother <- function(model, term = 1) {
  # mgcv::plot.gam invisibly returns the data used to construct the plot
  dat <- plot(model, select = 0)[[term]]
  
  ndim <- sum(c("x", "y") %in% names(dat))
  
  fit <- dat$fit[,1]
  lwr <- fit - 2 * dat$se
  upr <- fit + 2 * dat$se
  
  if (ndim == 1) {
    out <- data.frame(x = dat$x, fit, lwr, upr)
  }
  else if (ndim == 2) {
    xy <- expand.grid(dat$x, dat$y)
    out <- data.frame(x = xy[,1], y = xy[,2], fit, lwr, upr)
  }
  else
    stop("Only 1D and 2D smoothers are supported")

  out
}