Functional residual density plot

Generates a functional residual-vs-predictor plot, visualizing the density of the functional residuals as a heatmap rendered on either the uniform or normal scale.

Usage

fredplot(
  object,
  x,
  resolution = 101,
  scale = c("uniform", "normal"),
  type = c("kde", "hex"),
  n = 100,
  plot = TRUE,
  color.palette = function(n) hcl.colors(n, "YlGnBu"),
  colorkey = FALSE,
  smooth = TRUE,
  smooth.col = "white",
  smooth.lwd = 1,
  smooth.lty = "solid",
  xlab = deparse1(substitute(x)),
  ylab = "Residual density",
  ...
)

Arguments

object

An object for which the extraction of functional residuals is meangingful (e.g., a glm) object.

x

Vector of predictor values to use for the x-axis. Ideally, x should represent a continuous variable. Categorical variables should be coerced to integer. Using jitter() can also be useful. For instance, something like x = jitter(as.integer(df$x)) could be useful. If x is categorical in nature, it's also recommended to turn off LOESS smoothing by setting smooth = FALSE.

resolution

Integer specifying the resolution of the plot. Default is 101.

scale

Character string specifying which scale to use for plotting. Default is "uniform".

type

Character string specifying the type of plot to construct. Current options include "hex" (the default) for a hexagonal heatmap of 2D bin counts, or "kde" for 2D kernel density estimation.

n

Integer specifying the number of grid points in each direction for two-dimensional kernel density estimation; passed to kde2d() whenever type = "kde". Default is 100.

plot

Logical indicating whether to return a data frame that can be plotted manually (FALSE) or plot the results automatically (TRUE). Default is TRUE. Setting to FALSE is useful for debugging or advanced usage.

color.palette

A color palette function to be used to assign colors in the plot.

colorkey

Logical indicating whether or not a legend should be drawn. Default is FALSE.

smooth

Logical indiacting whether or not to include a LOESS smoother. Default is TRUE,

smooth.col

Integer or character string specifying the line color to use for the optional LOESS smoother; default is "white".

smooth.lwd

Integer specifying the line width to use for the optional LOESS smoother; default is 1.

smooth.lty

Integer or character string specifying the line type to use for the optional LOESS smoother; default is 1 (equivalent to "solid").

xlab, ylab

Character strings specifying the labels to use for the x- and y-axes, respectively.

...

Additional optional arguments passed to lattice::levelplot() (type = "kde") or hexbin::hexbinplot() (type = "hex").

Value

TBD.

Examples

# Generate data from a logistic regression model with quadratic form
set.seed(1217)
n <- 1000
x <- rnorm(n)
z <- 1 - 2*x + 3*x^2 + rlogis(n)
y <- ifelse(z > 0, 1, 0)

# Fit models with/without quadratic term
bad <- glm(y ~ x, family = binomial)  # wrong
good <- glm(y ~ x + I(x^2), family = binomial)  # right
#> Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

# Functional REsidual Density plot for each model
gridExtra::grid.arrange(
  fredplot(bad, x = x, type = "hex", aspect = 1),
  fredplot(good, x = x, type = "hex", aspect = 1),
  nrow = 1
)