Skip to contents

Function residuals

fresiduals.Rd

Computes the function residual described in TBD.

Usage

fresiduals(
  object,
  type = c("function", "surrogate", "probscale"),
  link.scale = TRUE,
  ...
)

Arguments

object

A fitted model object.

type

Character string specifying the type of residual to compute. Current options include:

  • "function" - (the default) for a list of functional residuals;

  • "surrogate" for a sample of surrogate residuals;

  • "probscale" for probability-scale residuals.

Logical indicating whether or not surrogate residuals (type = "surrogate") should be returned on the link scale (TRUE) vs. the probability scale (FALSE). Default is TRUE.

...

Additional optional arguments. Currently ignored.

Value

Either a list of functions (type = "function") that also inherits from class "funres" or a vector of residuals (type = "surrogate" or type = "probscale").

References

TODO: Reference paper when published in JASA.

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
fit.wrong <- glm(y ~ x, family = binomial)  # wrong
fit.right <- glm(y ~ x + I(x^2), family = binomial)  # right
#> Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

# Generate functional residuals
fres.wrong <- fresiduals(fit.wrong)
plot(fres.wrong[[1]])  # plot first functional residual


# Function-function plot
par(mfrow = c(1, 2))
ffplot(fres.wrong, type = "l")
ffplot(fit.wrong, type = "l")


# Residual vs. predictor plot for each model based on surrogate method
par(mfrow = c(1, 2), las = 1)
lpars <- list(col = 2, lwd = 2)
col <- adjustcolor(1, alpha.f = 0.1)
palette("Okabe-Ito")
scatter.smooth(x, y = fresiduals(fit.wrong, type = "surrogate"),
               lpars = lpars, col = col, main = "Wrong model",
               xlab = "x", ylab = "Surrogate residual")
abline(h = 0, col = 3, lty = 2)
scatter.smooth(x, y = fresiduals(fit.right, type = "surrogate"),
               lpars = lpars, col = col, main = "Correct model",
               xlab = "x", ylab = "Surrogate residual")
abline(h = 0, col = 3, lty = 2)