Thank you for helping improve lambertReg. Start with a GitHub issue describing the proposed change or a small reproducible example of the problem.
A useful report
Include the package and R versions, operating system, function call, controls, and the numerical status of the fit. Use a synthetic or openly licensed example and omit confidential data. Include a seed and foldid when the issue involves cross-validation. For plotting issues, include the plotting call and device size.
Scope
Documentation, examples, accessibility, and interface improvements are welcome. Changes to the estimator, preprocessing, scalar update, coordinate order, or acceptance criteria require an explicit methodological discussion. A faster implementation must preserve the relevant coefficients, CV scores, and diagnostics against the reference implementation; runtime alone is not sufficient evidence.
Local checks
From a clone’s parent directory, run:
Use the tests in tests/ and small examples. Do not attach large datasets or launch a production study as part of a routine contribution. Package examples use dense finite predictor matrices and Gaussian responses; preserve the current function signatures and defaults unless the change is agreed in advance.
Documentation site
The optional site build uses pkgdown, knitr, rmarkdown, and Pandoc. In R, from the repository directory:
pkgdown::build_site(preview = FALSE)The local site is written to docs/. Inspect the introduction, reference pages, and example plots on both desktop and a narrow screen before proposing changes.