Skip to contents

Add a Linear Regression Line

Source: R/lines.lm.R
linesLm.Rd

Add a linear regression line to an existing plot. The function first calculates predictions from an lm object and then adds the fitted line together with optional confidence and prediction bands.

Usage

# S3 method for class 'lm'
lines(
  x,
  col = pal()[1],
  lwd = 2,
  lty = "solid",
  type = "l",
  n = 100,
  cband = list(conf.level = 0.95),
  pband = NA,
  xpred = NULL,
  ...
)

# S3 method for class 'lmlog'
lines(
  x,
  col = pal()[1],
  lwd = 2,
  lty = "solid",
  type = "l",
  n = 100,
  cband = list(conf.level = 0.95),
  pband = NA,
  xpred = NULL,
  ...
)

Arguments

x

linear model object as returned by lm().

col

line color. Defaults to pal()[1].

lwd

line width.

lty

line type.

type

plotting type passed to lines().

n

number of points used for plotting the fit.

cband

controls the confidence band. May be TRUE, FALSE, NULL, NA, or a named list. The confidence level is specified via conf.level, all other elements are graphical parameters of the band. Default is list(conf.level=0.95).

pband

controls the prediction band, as cband. Default is NA.

xpred

optional numeric vector defining the range over which predictions should be calculated.

...

further graphical parameters passed to lines() for the fitted line.

Value

No return value; called for its side effect.

Details

In contrast to abline(), polynomial models and transformed predictors are supported as long as the model contains exactly one predictor variable. A polynomial, y ~ poly(x, 2) or y ~ x + I(x^2), is drawn against x; a single transformed term such as y ~ log(x) is drawn against the transformed values, matching plot(y ~ log(x)).

Confidence and prediction bands are controlled via cband and pband. These arguments can be:

  • FALSE, NULL or NA: suppress the band

  • TRUE: draw the band with default settings

  • a named list: customize the band appearance and confidence level

See also

lines(), lm()

Other graphics.trendlines: lines.loess(), splineCI

Examples

plot(dist ~ speed, cars)
lines(lm(dist ~ speed, cars))


# polynomial model with a 99% confidence band and a prediction band
plot(dist ~ speed, cars)
lines(lm(dist ~ poly(speed, 2), cars), col = "red",
      cband = list(conf.level = 0.99), pband = TRUE)


# transformed predictor, no bands
plot(mpg ~ log(hp), mtcars)
lines(lm(mpg ~ log(hp), mtcars), cband = FALSE)