Harden pcatR summaries, suppression, and PDF export before CRAN

[JaeManP]

Summary

This draft hardens the pcatR 1.0.1 pre-CRAN candidate by correcting:

• the complete five-category neutral numerator through a distinct n_neutral_complete;
• the five-category partition invariant by deriving all five counts from pcat_class5;
• pct_complete_class as the share of valid-direction records with a complete five-category classification;
• small-cell suppression so modal_class is hidden with numeric analytic results;
• unavailable polarization so it is represented as NA, not confirmed FALSE;
• invalid-direction strength consistency;
• failure-safe, preflighted profile-PDF export, including directory-path protection and finite page dimensions;
• finite scalar validation for consensus and action-plan probability thresholds;
• source-archive README links to excluded release records; and
• provisional release-validation documentation.

Why

The previous implementation could produce a complete-category distribution above 100%, expose a modal response classification in a suppressed cell, assign a valid-looking strength to an invalid direction, leave a partial PDF after a later-page failure, unintentionally accept vector threshold inputs, document a completeness field that was absent from the API, and potentially rename an existing directory supplied as an overwrite target.

Methodological boundaries

• No item wording changed.
• No original or updated CFIR mapping changed.
• No response-code meaning changed.
• No purported validated overall pCAT score was introduced.
• Directional and complete-category denominators remain intentionally distinct.
• Strategy candidates remain deliberation prompts rather than prescriptions.

Repository integration

• PR Record Lilac Li CRAN authorization #3 was merged by Lilac Li before this branch was restacked.
• Current base branch: main.
• Current base and merge-base SHA: c1a93994633a94608238565e8f893a797d8e32a7.
• Current hardening head: 43ae025a491ce6ebde713b1431588a983155c83b.
• The restacked tree is byte-for-byte identical to the previously validated hardening tree; only commit identities changed during rebase.
• Previous workflow, artifact, and candidate-hash identifiers are historical and are not evidence for this restacked head.

Local verification

• devtools::test(stop_on_failure = TRUE): 236 expectations, 0 failures, 0 errors, 0 warnings, 0 skips.
• Focused summary/consensus and analysis/export regression files: 139 expectations, all passing.
• pcat_self_test(): 25/25 passing.
• devtools::check(cran = FALSE, error_on = "never"): 0 errors, 0 warnings, 0 notes.
• Exact local archive: pcatR_1.0.1.tar.gz.
• Pre-freeze local archive SHA-256: a8ee1f3e02861eda11281e4390207335339dd4312cf17ab69bf4a925eee44377.
• Archive entries: 81.
• Archive inspection: guide source/HTML/PDF present; private, development, and release-control files absent; archived README has no relative links to excluded release records.
• Clean exact-archive installation: passed; version 1.0.1, 25 self-tests, analytical regressions, suppression, successful PDF export, failed-export cleanup, and directory-target preservation all passed.
• R CMD check --as-cran on the exact local archive: 0 errors, 0 warnings, 3 explained notes (new submission, host clock unavailable, and the local R 4.4 validator rejecting R-generated HTML5 elements).
• Spelling: 0 findings.
• URL check: all 14 URLs correct.
• Clean pkgdown build: passed; home, reference, articles, citation, authors, license, attribution, release, and news pages inspected for the amended content.
• Guide render: self-contained HTML and 24-page Letter PDF regenerated; every PDF page visually inspected with no clipping, overlap, malformed tables, missing glyphs, or broken footers.

Current GitHub Actions verification

Exact restacked head under test: 43ae025a491ce6ebde713b1431588a983155c83b.

• R CMD check workflow: 29672326904 — passed.
• Coverage workflow: 29672326903 — passed; overall line coverage: 84.86%.
• pkgdown workflow: 29672326930 — passed; deployment was intentionally skipped for this draft PR.
• Passed standard matrix jobs: Windows R-release, macOS R-release, Ubuntu R-release, R-devel, R-oldrel-1, and R 4.1.
• Exact archive build job: 88153490605.
• Exact downloaded-archive full-manual jobs: R-release 88153566507; R-devel 88153566514.
• Source artifact: ID 8437716422, name pcatR-1.0.1-source.
• Check-log artifacts: R-release ID 8437732735; R-devel ID 8437733655.
• pkgdown artifact: ID 8437718938.
• CI candidate archive: pcatR_1.0.1.tar.gz.
• CI candidate archive SHA-256: 61c549ae6ebc764f0471bda85cae7cccf710276b4a4ef99053afe7b7059b79b9.
• The internal pcatR_1.0.1.sha256 agrees with the independently downloaded archive.
• Both exact-archive jobs verified the checksum before checking the same bytes.
• Both exact-archive checks completed with 0 errors, 0 warnings, and 2 explained notes: new submission and unavailable HTML Tidy.
• Independent inspection confirmed 81 archive entries; all guide formats and amended source code are present; the private statement, release-control files, and development-only files are absent; and the archived README uses explicit repository links for excluded release records.
• The downloaded pkgdown artifact contains the new completeness field, unknown-polarization semantics, PDF path/dimension safeguards, and correct Lilac Li/Jae Man Park authorship order.

The completed pre-restack workflow identifiers and artifacts have been superseded by this exact-head run set and are not presented as current evidence.

Remaining external gates

• Original instrument authors' written package-name and source-content confirmation.
• Final source freeze.
• Exact frozen-archive Win-builder R-release and R-devel checks.
• CRAN submission and email confirmation.
• Post-acceptance v1.0.1 tag and GitHub Release.

This PR does not submit the package to CRAN, does not run or claim Win-builder validation, does not freeze the final submission archive, and does not create a release tag.