Residuals from a model fit

stat_fit_residuals fits a linear model and returns residuals ready to be plotted as points.

Usage

stat_fit_residuals(
  mapping = NULL,
  data = NULL,
  geom = "point",
  method = "lm",
  method.args = list(),
  n.min = 2L,
  formula = NULL,
  resid.type = NULL,
  weighted = FALSE,
  position = "identity",
  na.rm = FALSE,
  orientation = NA,
  show.legend = FALSE,
  inherit.aes = TRUE,
  ...
)

Arguments

mapping: The aesthetic mapping, usually constructed with aes. Only needs to be set at the layer level if you are overriding the plot defaults.
data: A layer specific dataset - only needed if you want to override the plot defaults.
geom: The geometric object to use display the data
method: function or character If character, "lm", "rlm", "rq" and the name of a function to be matched, possibly followed by the fit function's method argument separated by a colon (e.g. "rq:br"). Functions implementing methods must accept arguments to parameters formula, data, weights and method. A residuals() method must exist for the returned model fit object class.
method.args: named list with additional arguments.
n.min: integer Minimum number of distinct values in the explanatory variable (on the rhs of formula) for fitting to the attempted.
formula: a "formula" object. Using aesthetic names instead of original variable names.
resid.type: character passed to residuals() as argument for type (defaults to "working" except if weighted = TRUE when it is forced to "deviance").
weighted: logical If true weighted residuals will be returned.
position: The position adjustment to use for overlapping points on this layer
na.rm: a logical indicating whether NA values should be stripped before the computation proceeds.
orientation: character Either "x" or "y" controlling the default for formula.
show.legend: logical. Should this layer be included in the legends? NA, the default, includes if any aesthetics are mapped. FALSE never includes, and TRUE always includes.
inherit.aes: If FALSE, overrides the default aesthetics, rather than combining with them. This is most useful for helper functions that define both data and aesthetics and should not inherit behaviour from the default plot specification, e.g. borders.
...: other arguments passed on to layer. This can include aesthetics whose values you want to set, not map. See layer for more details.

Details

This stat can be used to automatically plot residuals as points in a plot. At the moment it supports only linear models fitted with function lm() or rlm(). It applies to the fitted model object methods residuals or weighted.residuals depending on the argument passed to parameter weighted.

A ggplot statistic receives as data a data frame that is not the one passed as argument by the user, but instead a data frame with the variables mapped to aesthetics. In other words, it respects the grammar of graphics and consequently within the model formula names of aesthetics like $x$ and $y$ should be used instead of the original variable names, while data is automatically passed the data frame. This helps ensure that the model is fitted to the same data as plotted in other layers.

Note

How weights are applied to residuals depends on the method used to fit the model. For ordinary least squares (OLS), weights are applied to the squares of the residuals, so the weighted residuals are obtained by multiplying the "deviance" residuals by the square root of the weights. When residuals are penalized differently to fit a model, the weighted residuals need to be computed accordingly. Say if we use the absolute value of the residuals instead of the squared values, weighted residuals are obtained by multiplying the residuals by the weights.

Computed variables

Data frame with same value of nrow as data as subset for each group containing five numeric variables.

x: x coordinates of observations or x residuals from fitted values
y: y coordinates of observations or y residuals from fitted values
x.resid: residuals from fitted values
y.resid: residuals from fitted values
weights: the weights passed as input to lm or those computed by rlm

For orientation = "x", the default, stat(y.resid) is copied to variable y, while for orientation = "y" stat(x.resid) is copied to variable x.

Examples

# generate artificial data
set.seed(4321)
x <- 1:100
y <- (x + x^2 + x^3) + rnorm(length(x), mean = 0, sd = mean(x^3) / 4)
my.data <- data.frame(x, y)

# plot residuals from linear model
ggplot(my.data, aes(x, y)) +
  geom_hline(yintercept = 0, linetype = "dashed") +
  stat_fit_residuals(formula = y ~ x)


ggplot(my.data, aes(x, y)) +
  geom_hline(yintercept = 0, linetype = "dashed") +
  stat_fit_residuals(formula = y ~ x, weighted = TRUE)


# plot residuals from linear model with y as explanatory variable
ggplot(my.data, aes(x, y)) +
  geom_vline(xintercept = 0, linetype = "dashed") +
  stat_fit_residuals(formula = x ~ y) +
  coord_flip()


# give a name to a formula
my.formula <- y ~ poly(x, 3, raw = TRUE)

# plot residuals from linear model
ggplot(my.data, aes(x, y)) +
  geom_hline(yintercept = 0, linetype = "dashed") +
  stat_fit_residuals(formula = my.formula) +
  coord_flip()


ggplot(my.data, aes(x, y)) +
  geom_hline(yintercept = 0, linetype = "dashed") +
  stat_fit_residuals(formula = my.formula, resid.type = "response")


# plot residuals from robust regression
ggplot(my.data, aes(x, y)) +
  geom_hline(yintercept = 0, linetype = "dashed") +
  stat_fit_residuals(formula = my.formula, method = "rlm")


# plot residuals with weights indicated by colour
my.data.outlier <- my.data
my.data.outlier[6, "y"] <- my.data.outlier[6, "y"] * 10
ggplot(my.data.outlier, aes(x, y)) +
  stat_fit_residuals(formula = my.formula, method = "rlm",
                      mapping = aes(colour = after_stat(weights)),
                      show.legend = TRUE) +
  scale_color_gradient(low = "red", high = "blue", limits = c(0, 1),
                       guide = "colourbar")


# plot weighted residuals with weights indicated by colour
ggplot(my.data.outlier) +
  stat_fit_residuals(formula = my.formula, method = "rlm",
                     mapping = aes(x = x,
                                   y = stage(start = y, after_stat = y * weights),
                                   colour = after_stat(weights)),
                     show.legend = TRUE) +
  scale_color_gradient(low = "red", high = "blue", limits = c(0, 1),
                       guide = "colourbar")


# plot residuals from quantile regression (median)
ggplot(my.data, aes(x, y)) +
  geom_hline(yintercept = 0, linetype = "dashed") +
  stat_fit_residuals(formula = my.formula, method = "rq")


# plot residuals from quantile regression (upper quartile)
ggplot(my.data, aes(x, y)) +
  geom_hline(yintercept = 0, linetype = "dashed") +
  stat_fit_residuals(formula = my.formula, method = "rq",
  method.args = list(tau = 0.75))


# inspecting the returned data
gginnards.installed <- requireNamespace("gginnards", quietly = TRUE)

if (gginnards.installed)
  library(gginnards)

if (gginnards.installed)
  ggplot(my.data, aes(x, y)) +
   stat_fit_residuals(formula = my.formula, resid.type = "working",
                      geom = "debug")

#> [1] "PANEL 1; group(s) -1; 'draw_function()' input 'data' (head):"
#>   x          y    y.resid x.resid weights PANEL group orientation
#> 1 1 -23513.687 -23513.687      NA       1     1    -1          NA
#> 2 2 -11663.732 -11663.732      NA       1     1    -1          NA
#> 3 3  47289.460  47289.460      NA       1     1    -1          NA
#> 4 4  54175.231  54175.231      NA       1     1    -1          NA
#> 5 5  -8620.676  -8620.676      NA       1     1    -1          NA
#> 6 6 101247.937 101247.937      NA       1     1    -1          NA

if (gginnards.installed)
  ggplot(my.data, aes(x, y)) +
    stat_fit_residuals(formula = my.formula, method = "rlm",
                       geom = "debug")

#> [1] "PANEL 1; group(s) -1; 'draw_function()' input 'data' (head):"
#>   x          y    y.resid x.resid   weights PANEL group orientation
#> 1 1 -24689.771 -24689.771      NA 1.0000000     1    -1          NA
#> 2 2 -12710.840 -12710.840      NA 1.0000000     1    -1          NA
#> 3 3  46364.845  46364.845      NA 1.0000000     1    -1          NA
#> 4 4  53366.731  53366.731      NA 1.0000000     1    -1          NA
#> 5 5  -9319.337  -9319.337      NA 1.0000000     1    -1          NA
#> 6 6 100652.946 100652.946      NA 0.7800764     1    -1          NA