README.md
In mfrmr: Estimation and Diagnostics for Many-Facet Measurement Models

mfrmr validation artifacts

This directory contains non-exported release-review helpers and evidence artifacts. They are included with the package so that release decisions can be reconstructed from source files, check logs, and documented validation criteria.

These files are not user-facing analysis functions. They support release review, CRAN submission preparation, and future maintenance. Public release notes stay in NEWS.md; implementation-level evidence, source-grounding, and long-run validation details belong here.

release-readiness.R: release-readiness review. Source this file and run mfrmr_release_readiness_review(pkg_dir = ".") from the package root. The review checks version labels, the local check log, the CI check workflow, public terminology, and the release evidence files.
release-evidence-map-0.2.0.md: narrative review map linking release claims to mathematical, statistical, UX, documentation, and engineering evidence.
release-evidence-map-0.2.1.md: source-grounded evidence map for the 0.2.1 bounded-GPCM recovery-review refinements, including the boundary between cited model literature and package-specific validation labels.
release-evidence-checklist-0.2.1.csv: structured checklist used by the readiness helper and by manual release review for the current release. The 0.2.0 checklist is retained as historical release evidence.
gpcm-post-0.2.1-roadmap.md: maintenance roadmap for bounded-GPCM surfaces that remain caveated, blocked, or deferred after 0.2.1, including score-side review, report/QC bundles, design and screening operating characteristics, linking synthesis, posterior predictive checks, and heavy-backend extensions.
external-parameter-recovery-simulation-0.2.0.md: compact review of the separate common-data parameter-recovery simulation workflow. The large generated datasets and engine outputs are not bundled with the package; this file records the release-relevant evidence and its limits.
external-recovery-audit.R: optional audit helper that reads a local Parameter_Recovery_Simulation/ output directory, checks expected CSV schemas, records file fingerprints, and regenerates the compact evidence summary tables used for release review.

Run these commands from the package root after any source, roxygen, vignette, or compiled-code change:

R CMD build .
R CMD check --no-manual --as-cran mfrmr_0.2.1.tar.gz

Then run:

source("inst/validation/release-readiness.R")
readiness <- mfrmr_release_readiness_review(pkg_dir = ".")
summary(readiness)

The release candidate should have Status: OK in the local check log and no concern rows in readiness$gate_summary. The check log must also report the same package version as DESCRIPTION; stale logs from an earlier release are reported as a package-check concern. If the local environment cannot verify external clock time, record that environment-only NOTE in cran-comments.md and rerun the package check with the clock check disabled to confirm that package checks are otherwise clean.

CRAN-time tests are intentionally lightweight because CRAN check hosts have strict timing constraints. Run the full non-CRAN regression surface separately when release evidence is needed:

NOT_CRAN=true Rscript -e 'testthat::test_local(".")'

If the external common-data simulation workflow has been refreshed, audit it from the package side before updating the evidence summary:

source("inst/validation/external-recovery-audit.R")
external_review <- mfrmr_review_external_recovery_simulation(
  "../Parameter_Recovery_Simulation"
)
summary(external_review)

source("inst/validation/release-readiness.R")
readiness <- mfrmr_release_readiness_review(
  pkg_dir = ".",
  external_recovery_dir = "../Parameter_Recovery_Simulation"
)
summary(readiness)$external_recovery_status

GitHub Actions runs the package on macOS, Windows, and Linux across release, oldrel, and devel R. Warnings are treated as check failures. The workflow also uploads the check directory as an artifact for each matrix job so that release review can compare local and CI evidence instead of relying only on the final job status.

The readiness helper checks the workflow contract from source. It does not replace reading the uploaded CI artifacts before release submission. The external parameter-recovery summary is an additional source-grounded review artifact, not a substitute for rerunning the package tests or the optional long-running validation scripts.