qresid

Randomized quantile residuals for regression diagnostics in Stata

qresid implements randomized quantile residuals (Dunn-Smyth residuals) for regression diagnostics in supported independent regression models in Stata. The package is intended for generalized linear and related models where conventional residuals may be difficult to interpret.

In practice, qresid gives applied users one graphical language for a broad range of fitted probability models: Gaussian and positive continuous GLMs, Bernoulli and binomial responses, Poisson and negative-binomial counts, and selected extensions for zero inflation, truncation, censoring, hurdle structures, frequency weights, and fixed-dispersion CDF replay.

Installation

qresid is available from SSC. The recommended installation command is:

ssc install qresid

After installation:

which qresid
help qresid

The GitHub repository is the public source, documentation, issue-tracking, and development home for the package. Users who want the current GitHub source can install with:

net install qresid, from("https://raw.githubusercontent.com/psotob91/qresid/main/") replace

Then check the installation in the same way:

which qresid
help qresid

Quick Start

Gaussian regression:

sysuse auto, clear
regress price mpg
qresid rq
qnorm rq
predict double muhat, xb
scatter rq muhat, yline(0)

Poisson regression:

sysuse auto, clear
poisson rep78 mpg if rep78 < .
qresid rq_pois, seed(12345)
qnorm rq_pois

Tutorials

Tutorial	What you will learn
Continuous outcomes	Compare Gaussian, Gamma, and inverse-Gaussian fits when positive continuous outcomes make ordinary residual plots hard to read.
Binary and binomial outcomes	Diagnose Bernoulli and binomial-count models on a normal-score residual scale instead of reading bands of Pearson or deviance residuals.
Link and functional-form checks	Detect link-function mistakes, nonlinear covariate effects, and seasonal patterns that can be subtle in binary or discrete data.
Count outcomes	Move beyond a generic "not Poisson" conclusion by comparing Poisson, negative-binomial, and generalized-Poisson alternatives graphically.
Zero-inflated, truncated, and censored counts	Separate lack of fit due to extra zeros, restricted support, censoring intervals, and count-tail behavior.
Hurdle count models	Compare hurdle and zero-inflated interpretations when zeros and positive counts may arise from different mechanisms.
Weights and dispersion replay	Understand how frequency weights and fixed-dispersion CDF replay affect residual diagnostics without changing the fitted model.

Additional documentation

Reference	What it covers
Manual index	Full tutorial sequence, reproduction notes, and links to generated figures and output excerpts.
Markdown reference	GitHub-friendly version of the installed Stata help, including syntax, options, stored results, and formulas.
Features and diagnostic workflow	Documented model classes and a concise workflow for Q-Q plots and residual-versus-fitted or covariate plots.
Statistical background	PIT/RQR formulas, special count-model CDFs, and methodological references.
Validation and benchmarking	Validation principles, independent benchmark references, and Stata 15.0 compatibility checks.
Supported specifications	Public support scope for this release, including weights, dispersion, and optional external estimators.
Validation evidence	Summary table of validation approaches and links to public test entry points.
Project scope	Current boundaries and future extensions.
News and changelog	Release notes and documentation changes.

Contributing

Contributions, bug reports, reproducible examples, and methodological discussion are welcome. Please read the contributing guide and code of conduct before opening a pull request.

Citation

If you use qresid in academic work, please cite:

Soto-Becerra, P. 2026. qresid: Quantile and randomized quantile residuals for Stata models. Version 1.0.0. Available from https://github.com/psotob91/qresid

Maintainer

Role	Name	Affiliation	Contact
Maintainer	Percy Soto-Becerra, MD, MSc, PhD(c)	Universidad Privada del Norte, Lima, Peru	percy.soto@upn.edu.pe; percys1991@gmail.com

License

MIT