Display a 2D slice of a linear model

geom_slice() draws the prediction line of a fitted lm() on a ggplot — a 2D slice of a possibly high-dimensional model. It looks like ggplot2::geom_smooth(), but where geom_smooth() fits its own model to the plotted data, geom_slice() draws your model. The predictor mapped to the plot's x-axis varies along the line; every other predictor is fixed, and geom_slice() reports how, so it is always clear which slice of the model you are looking at:

Usage

geom_slice(
  model,
  n = 100,
  inherit.aes = TRUE,
  predict_vars = list(),
  back_transform = NULL,
  interval = "none",
  band = FALSE,
  full_range = FALSE,
  ...,
  fullrange = NULL
)

Arguments

model

A linear model fitted by lm().

n

Number of prediction points along the line (default 100).

inherit.aes

If TRUE (default), inherit aesthetics from the ggplot() call.

predict_vars

A named list of values at which to hold predictors not shown on the plot, such as predict_vars = list(hp = 110). Unlisted predictors are imputed (with a message). Giving a variable several values draws one line per value, and several multi-value variables are crossed: predict_vars = list(x2 = c(1, 2, 3), x3 = c(1, 4)) draws 6 lines.

back_transform

How to map predictions onto the y-axis when the model's response is transformed. Default NULL (and TRUE) auto-detects from the model formula; FALSE turns back-transformation off; a one-argument function (e.g. exp) or a name ("log", "log10", "log2", "sqrt", "exp", "inverse") applies that transformation.

interval

Draw a ribbon around the line: "none" (default), or "confidence" / "prediction" for the corresponding predict.lm() interval. The ribbon follows its line's color unless you set fill. Cannot be combined with band.

band

Draw a projection band — two edge slices with a translucent ribbon between them — instead of a single line. band = "variable" spans that predictor: between the values you gave in predict_vars (e.g. predict_vars = list(x2 = c(1, 4))), or its observed data range when predict_vars leaves it out. band = TRUE infers the variable: the one multi-value predict_vars entry, or the single predictor the plot does not otherwise show. Default FALSE.

full_range, fullrange

If TRUE, each line spans the full x range of the panel instead of stopping at its group's own data range (like fullrange in ggplot2::geom_smooth(); fullrange is accepted as an alias). Default FALSE. Only applies when the x-axis maps a single predictor; with a composite x-axis expression predictions are made at the data points, so the lines keep their data extent.

...

Other arguments passed to the layer, such as fixed aesthetics (color = "red", linewidth = 1.2).

Value

A ggplot2 layer that draws the slice.

Details

• Variables named in predict_vars are held at your chosen values.

• Variables mapped to a grouping aesthetic (e.g. aes(color = g)) are pinned to each group's own value — one line per group. Any interval or band ribbon is drawn in its own line's color.

• Facet variables are pinned to each panel's value.

• Anything left over is imputed (mean for numeric, most common value for factor/character), with a console message naming the value used.

If the model's response is transformed (e.g. lm(log(y) ~ x)) but the plot shows raw y, predictions are automatically back-transformed to match the y-axis (a message says so). Axis expressions (aes(log(y)), aes(x * x2)) and transformed scales (scale_x_log10()) are also handled.

See also

• geom_slice_text() to label each line with the values that produced it.

• geom_slice_subtitle() / geom_slice_caption() to describe the slice (model equation, held values) in the plot's subtitle or caption.

• autoplot.lm() for a complete data-plus-slice plot in one call.

• ggplot2::geom_smooth(), which this layer resembles, except that it draws a model you fitted rather than fitting its own.

Examples

library(ggplot2)

# Basic use, like geom_smooth() but drawing *your* model. Sepal.Width is
# not on the plot, so it is held at its mean (a console message says so).
model <- lm(Sepal.Length ~ Sepal.Width + Petal.Length, data = iris)
ggplot(iris, aes(Petal.Length, Sepal.Length)) +
  geom_point() +
  geom_slice(model)
#> Value for `Sepal.Width` not specified - Used mean: 3.057
#>     To choose a slice, use 'predict_vars = list(Sepal.Width = 3.057)'.


# Several values draw one line per value (label them with
# geom_slice_text()); several multi-value variables are crossed
ggplot(iris, aes(Petal.Length, Sepal.Length)) +
  geom_point() +
  geom_slice(model, predict_vars = list(Sepal.Width = c(2, 3, 4))) +
  geom_slice_text()


# Confidence or prediction ribbon around the line
model2 <- lm(hwy ~ displ + I(displ^2), data = mpg)
ggplot(mpg, aes(displ, hwy)) +
  geom_point() +
  geom_slice(model2, interval = "prediction")


# A grouping aesthetic pins its predictor to each group's own value:
# one line per drive type, each stopping at the end of its group's data
model3 <- lm(hwy ~ displ + I(displ^2) + displ:drv, data = mpg)
ggplot(mpg, aes(displ, hwy, color = drv)) +
  geom_point() +
  geom_slice(model3)


# full_range = TRUE extends each group's line across the whole panel
ggplot(mpg, aes(displ, hwy, color = drv)) +
  geom_point() +
  geom_slice(model3, full_range = TRUE)


# Facet variables pin per panel the same way
ggplot(mpg, aes(displ, hwy)) +
  geom_point() +
  geom_slice(model3) +
  facet_wrap(~drv)


# A projection band shows the reach of a predictor, here cyl from 5 to 7;
# band = TRUE infers the variable and its data range when it can
model4 <- lm(hwy ~ displ * drv + cyl, data = mpg)
ggplot(mpg, aes(displ, hwy, color = drv)) +
  geom_point() +
  geom_slice(model4, band = "cyl", predict_vars = list(cyl = c(5, 7)))


# A transformed response is back-transformed to match the y-axis: the
# model predicts log(mpg), the line appears in raw mpg units
model5 <- lm(log(mpg) ~ disp + hp, data = mtcars)
ggplot(mtcars, aes(disp, mpg)) +
  geom_point() +
  geom_slice(model5)
#> 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'.


# ... adapts when the axis itself is transformed
ggplot(mtcars, aes(disp, log(mpg))) +
  geom_point() +
  geom_slice(model5)
#> Value for `hp` not specified - Used mean: 146.7
#>     To choose a slice, use 'predict_vars = list(hp = 146.7)'.


# ... or control the mapping yourself with back_transform
ggplot(mtcars, aes(disp, mpg)) +
  geom_point() +
  geom_slice(model5, back_transform = exp)
#> Value for `hp` not specified - Used mean: 146.7
#>     To choose a slice, use 'predict_vars = list(hp = 146.7)'.


# n controls how many prediction points make up the line
ggplot(mtcars, aes(disp, mpg)) +
  geom_point() +
  geom_slice(model5, n = 10)
#> 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'.