Why do scatter.smooth from package stats and geom_smooth from package ggplot2 give different results?

Question

This is the loess smoothing curve in base r

dat <- data.frame(RT = c(151, 160, 168, 169, 172, 175, 189, 279),
              IQ = c(123, 121, 115, 110, 105, 103, 100, 120))
with(dat, scatter.smooth(IQ, RT, span = 0.8))

And this is the loess curve produced with ggplot:

ggplot(dat, aes(x = IQ, y = RT)) +
geom_point() +
geom_smooth(method = stats::loess, se = F, span = 0.8)

Why are they different?

Answer 1

It's because scatter.smooth() does not use the default loess() arguments, whereas geom_smooth() does. If you provide the method arguments that are default in scatter.smooth() to the ggplot2 method, they produce very similar results.

dat <- data.frame(RT = c(151, 160, 168, 169, 172, 175, 189, 279),
                  IQ = c(123, 121, 115, 110, 105, 103, 100, 120))
with(dat, scatter.smooth(IQ, RT, span = 0.8))

library(ggplot2)

ggplot(dat, aes(x = IQ, y = RT)) +
  geom_point() +
  geom_smooth(formula = y ~ x, method = "loess", se = F, span = 0.8,
              method.args = list(degree = 1, family = "symmetric"))

^{Created on 2021-02-03 by the reprex package (v1.0.0)}