plotSimDiags: Diagnostic plots for simulated data

Description

This provides diagnostic plots, closely equivalent to those provided by plot.lm, for simulated data. By default, simulated data are for the fitted model. Alternatively, simulated data can be supplied, making it possible to check the effct of fitting, e.g., an AR1 model.

Usage

plotSimDiags(obj, simvalues = NULL, seed = NULL,
types = NULL, which = c(1:3, 5), layout = c(4, 1), qqline=TRUE,
cook.levels = c(0.5, 1), caption = list("Residuals vs Fitted",
"Normal Q-Q", "Scale-Location", "Cook's distance", "Residuals vs Leverage",
expression("Cook's dist vs Leverage  " * h[ii]/(1 - h[ii]))),
...)

Value

A list of lattice graphics objects is returned, one for each value of

which. List elements for which a graphics object is not returned are set to NULL. Or if which is of length 1, a lattice graphics object.

residVSfitted: Residuals vs fitted
normalQQ: Normal quantile-quantile plot
scaleVSloc: Scale versus location
CookDist: Cook's distance vs observation number
residVSlev: Standardized residuals (for GLMs, standardized Pearson residuals) vs leverage
CookVSlev: Cook's distance vs leverage

For the default which=c(1:3,5), list items 1, 2, 3 and 5 above contain graphics objects, with list elements 4 and 6 set to NULL.

Arguments

obj: Fitted model object - lm or an object inheriting from lm
simvalues: Optional matrix of simulated data.
seed: Random number seed - set this to make results repeatable.
types: If set, this should be a list with six elements, ordinarily with each list element either "p" or c("p","smooth") or (which=2, which=6) NULL or (which=4) "h"
which: Set to be a subset of the numbers 1 to 6, as for plot.lm
layout: Controls the number of simulations and the layout of the plots. For example layout=c(3,4) will give 12 plots in a 3 by 4 layout.
qqline: logical: add line to normal Q-Q plot
cook.levels: Levels of Cook's statistics for which contours are to be plotted.
caption: list: Captions for the six graphs
...: Other parameters to be passed to plotting functions

Author

John Maindonald, with some code chunks adapted from plot.lm

Details

Diagnotic plots from repeated simulations from the fitted model provide a useful indication of the range of variation in the model diagnistics that are consistent with the fitted model.

References

See plot.lm

Examples

Run this code

women.lm <- lm(height ~ weight, data=women)
gphlist <- plotSimDiags(obj=women.lm, which=c(1:3,5))

Run the code above in your browser using DataLab