Postestimation diagnostic plots after GLM

[Dan Palmer]

The Stata Manual gives seven different options for postestimation diagnostic plots following the regress command:

• rvfplot - residual-versus-fitted plot
• avplot - added-variable plot
• avplots - all added-variables plots in one image
• cprplot - component-plus-residual plot
• acprplot - augmented component-plus-residual plot
• rvpplot - residual-versus predictor plot
• lvr2plot - leverage-versus-squared-residual plot

The manual does not, however, seem to have an equivalent list of posteestimation diagnostic plots following the glm command. I believe some of the above plots are appropriate in a GLM context, while I suspect others are not. Can anyone point me to Stata (or other) materials that might provide some instruction on which of the above commands would be appropriate following GLM? I'm also wondering if the suitability of certain plots would depend on the flavor of GLM used (e.g., family, link).

[Nick Cox]

The modeldiag package offers some possibilities:

SJ-10-1 gr0009_1 . . . . Software update for model diagnostic graph commands
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . N. J. Cox
(help anovaplot, indexplot, modeldiag, ofrtplot, ovfplot,
qfrplot, racplot, rdplot, regplot, rhetplot, rvfplot2,
rvlrplot, rvpplot2 if installed)
Q1/10 SJ 10(1):164
provides new command rbinplot for plotting means or medians
of residuals by bins; provides new options for smoothing
using restricted cubic splines; updates anova examples

SJ-4-4 gr0009 . . . . . . . . . . Speaking Stata: Graphing model diagnostics
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . N. J. Cox
(help anovaplot, indexplot, modeldiag, ofrtplot, ovfplot,
qfrplot, racplot, rdplot, regplot, rhetplot, rvfplot2,
rvlrplot, rvpplot2 if installed)
Q4/04 SJ 4(4):449--475
plotting diagnostic information calculated from residuals
and fitted values from regression models with continuous
responses

Note that there is a clash of names here: rdplot was introduced in this package in 2004, but the same name was used for a very different command in a later package for analysis of regression discontinuities. Sometime I will get around to changing the name. I was there first, but the other one has been widely noticed, and as an Editor of the SJ I bear part of the collective responsibility for telling the other authors that the name was in use. (They should have checked too, but still a little deal from all perspectives.)

The 2004 paper is accessible at https://www.stata-journal.com/sjpdf....iclenum=gr0009

But let's underline a simple principle. Most of these plots boil down to looking at residuals and fitted values in some way, so once you've used predict you can fire up your own plots quite easily.

[Dan Palmer]

Thank you Nick. I will put your 2004 article at the top of my reading list.