Automatically plot a linear model

autoplot() turns a fitted lm() into a complete ggplot in one call: the data the model was fitted to as a scatter, with a geom_slice() prediction line — and, by default, its confidence ribbon — through it. The result is a regular ggplot, so it can be extended with + as usual (labels, scales, more layers).

Usage

# S3 method for class 'lm'
autoplot(
  object,
  mapping = NULL,
  type = c("2d", "3d"),
  summary = TRUE,
  interval = "confidence",
  ...
)

Arguments

object: A linear model fitted by lm() with a data argument — the data is recovered from the model, so lm(y ~ x, data = your_data) works but lm(your_data$y ~ your_data$x) does not.
mapping: Aesthetics created with ggplot2::aes(), passed to ggplot2::ggplot() — for example mapping = aes(color = factor(cyl)). They resolve against the model's full data frame, so a column the model never mentions (such as cyl above) still works. An entry named x chooses the x-axis (mapping = aes(x = hp)), overriding the default first numeric predictor; an entry named y overrides the response axis.
type: "2d" (default) for a geom_slice() ggplot, or "3d" for an interactive scatter_3d() surface (needs exactly two numeric predictors).
summary: Whether to print summary(object) to the console before returning the plot (default TRUE). Printing it up front means you can refit the model and inspect its coefficients from inside the same autoplot() call; pass summary = FALSE to suppress it.
interval: The ribbon drawn around the slice line, passed to geom_slice(): "confidence" (default) for the mean's confidence interval, "prediction" for a single new observation's interval, or "none" for a bare line. Ignored when type = "3d".
...: Passed on to geom_slice() — for example predict_vars = list(hp = 110) to choose the slice, or fixed aesthetics such as color = "red" — or to scatter_3d() when type = "3d".

Value

A ggplot of the model's data with a slice of the model drawn through it, or — when type = "3d" — a plotly surface from scatter_3d().

Details

Pass type = "3d" for a model with exactly two numeric predictors to get an interactive scatter_3d() surface instead of the 2-D slice; ... is then forwarded to scatter_3d() (e.g. n, colors).

The model's first numeric predictor goes on the x-axis (override with mapping = aes(x = ...)) and the raw response variable goes on the y-axis. A predictor that groups into a few discrete lines rather than varying continuously — non-numeric, or numeric with at most five distinct values (like cyl's 4/6/8) — is worth showing, not holding at one value: the first such predictor not already on the plot becomes aes(color = ...) (one slice line per level) and a second one facets the plot. Everything past that is handled by geom_slice() and reported on the console: predictors not visible on the plot are imputed (mean for numeric, most common value for factor/character), and a transformed response such as lm(log(y) ~ x) is automatically back-transformed to match the raw y-axis.

Examples

library(ggplot2)

# A complete plot from a model in one call; summary() prints to the console
autoplot(lm(Petal.Length ~ Sepal.Length + I(Sepal.Length^2):Species, iris))
#> 
#> Call:
#> lm(formula = Petal.Length ~ Sepal.Length + I(Sepal.Length^2):Species, 
#>     data = iris)
#> 
#> Residuals:
#>      Min       1Q   Median       3Q      Max 
#> -0.72665 -0.15039  0.00562  0.16639  0.76639 
#> 
#> Coefficients:
#>                                     Estimate Std. Error t value Pr(>|t|)    
#> (Intercept)                         -3.79412    1.09932  -3.451 0.000731 ***
#> Sepal.Length                         2.04477    0.36389   5.619 9.51e-08 ***
#> I(Sepal.Length^2):Speciessetosa     -0.19763    0.02958  -6.680 4.75e-10 ***
#> I(Sepal.Length^2):Speciesversicolor -0.11503    0.03033  -3.793 0.000218 ***
#> I(Sepal.Length^2):Speciesvirginica  -0.09427    0.02987  -3.156 0.001944 ** 
#> ---
#> Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
#> 
#> Residual standard error: 0.262 on 145 degrees of freedom
#> Multiple R-squared:  0.9786,	Adjusted R-squared:  0.978 
#> F-statistic:  1655 on 4 and 145 DF,  p-value: < 2.2e-16
#> 
#> Coloured the plot by `Species` - one slice line per level.
#>     To choose the colour, use 'mapping = aes(color = ...)'.


# Predictors that group into a few values become color, then facets
model <- lm(hwy ~ displ + drv + cyl, data = mpg)
autoplot(model, summary = FALSE)
#> Coloured the plot by `drv` - one slice line per level.
#>     To choose the colour, use 'mapping = aes(color = ...)'.
#> Faceted the plot by `cyl` - one panel per level.
#>     For a single panel, drop `cyl` from the model.


# Change the defaults like a normal ggplot; the rest is still inferred
autoplot(model, aes(color = factor(cyl)), summary = FALSE) + facet_wrap(~drv)
#> Faceted the plot by `drv` - one panel per level.
#>     For a single panel, drop `drv` from the model.


# Choose the x-axis and pass geom_slice() options through
autoplot(model,
         aes(x = cyl, color = factor(drv)),
         interval = "none", summary = FALSE) +
  labs(title = "The Same Model From a Different Angle")
#> Value for `displ` not specified - Used mean: 3.472
#>     To choose a slice, use 'predict_vars = list(displ = 3.472)'.


# Transformed response, back-transformed onto the raw mpg axis
autoplot(lm(log(mpg) ~ disp + hp, data = mtcars), summary = FALSE)
#> Value for `hp` not specified - Used mean: 146.7
#>     To choose a slice, use 'predict_vars = list(hp = 146.7)'.
#> Predictions of `log(mpg)` were back-transformed to match the `mpg` axis.
#>     To turn this off, use 'back_transform = FALSE'.


# An interactive 3-D surface for a model with two numeric predictors
if (interactive()) {
  autoplot(lm(mpg ~ wt + I(wt^2) + hp, data = mtcars), type = "3d")
}

Usage

Arguments

Value

Details

See also

Examples