Add POOR

[strengejacke]

No description provided.

[Copilot]

Pull request overview

Adds and exports performance_poor() for logistic multilevel models, with tests, documentation, examples, and version updates.

Changes:

• Implements POOR calculations and print support.
• Adds tests and generated documentation.
• Updates MOR/IOR examples, NEWS, namespace, and package version.

Reviewed changes

Copilot reviewed 8 out of 11 changed files in this pull request and generated 4 comments.

Show a summary per file

File	Review summary
tests/testthat/test-performance_poor.R	Reviewed; no final comments.
tests/testthat/test-performance_mor.R	Reviewed; no final comments.
R/performance_poor.R	Critical: restrict calculations to logit models and handle random-slope or no-intercept models correctly. Nit: clarify the returned fraction versus percentage scale.
R/performance_mor.R	Reviewed; no final comments.
R/performance_ior.R	Reviewed; no final comments.
NEWS.md	Nit: correct rations to ratios.
NAMESPACE	Reviewed; no final comments.
man/performance_poor.Rd	Reviewed; no final comments.
man/performance_mor.Rd	Reviewed; no final comments.
man/performance_ior.Rd	Reviewed; no final comments.
DESCRIPTION	Reviewed; no final comments.

Files not reviewed (3)

• man/performance_ior.Rd: Generated file
• man/performance_mor.Rd: Generated file
• man/performance_poor.Rd: Generated file

Suppressed comments (2)

R/performance_poor.R:52

• The copied valid_ior name is incorrect for this function and makes the validation logic misleading. Rename it to valid_poor, including the conditional below.

  valid_ior <- .valid_roc_models(x) &&

R/performance_poor.R:72

• When both a fixed-effect estimate and a random-intercept variance are exactly zero, this computes 0 / sqrt(0) and returns NaN, contradicting the documented bounded POOR value. Zero random-effect variances are already possible in the singular fit used by this PR; define the zero/zero case explicitly and add a regression test.

        POOR = stats::pnorm(-abs(params$Estimate) / sqrt(2 * v_a[tau])),

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[DominiqueMakowski]

that's interesting, i was recently thinking of something similar, is there a reference to that? Can't this be generalized to any type of coefs not just odds ratios?

[strengejacke]

Look for merlo and logistic multilevel. One reference is https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0153778 (should add that to the function, too)