NEWS

projoint 1.1.1 (2026-02-22)

Fixed issue in organize_data() where profile-varying columns (e.g., selected_repeated) were incorrectly included in pivot_wider(id_cols), which could drop respondent-level covariates (e.g., weights) and cause .weights_2 to fail in choice_level estimation.
Improved cross-attribute pairing logic in choice-level QOI filtering.
Added regression tests for .weights_1 and .weights_2.

Fixed reshape_projoint()'' to handle repeated-task reshaping more robustly when tasks/outcomes are not in a simple 1..T order. The function now standardizes the mapping between the last outcome (the repeated measure) and the inferred base task, reducing misalignment in the task'' variable for some inputs.

Added stricter input/design validation in ``reshape_projoint()'':
- Checks that each base task contains exactly the two expected profiles (1 and 2) in the K-code design and aborts with an informative message otherwise.
- When ``.repeated = TRUE'', checks that each respondent provides the expected number of outcomes and aborts if repeated outcomes are missing.

Refactored the implementation of reshape_projoint()'' into a dedicated source file (R/reshape_projoint.R'') with no changes to the exported function name or arguments.

This release addresses all issues raised in the CRAN review and includes several small improvements and documentation updates.

DESCRIPTION: References now follow the required format authors (year) <https://...>.
Examples: Removed commented-out code and replaced examples of unexported functions (projoint_data()) with runnable toy examples.
Internal helpers: Suppressed documentation for unexported internal functions (projoint_tau(), plot_projoint_profile_level()) using @noRd.
Cross-references: Fixed a missing link warning in plot.projoint_results.Rd.
make_projoint_data(): Fixed a minor error affecting input validation.
Vignettes and website: Revised and expanded for clarity, examples, and reproducibility.

R CMD check --as-cran runs cleanly with 0 errors, 0 warnings, and 2 expected NOTES (package size and timestamps).

Revised pj_estimate() to automatically set the appropriate standard-error option (se_type) depending on clustering and estimation method.
Expanded documentation across functions:
- All exported methods now include \value{} sections describing return values.
- Clearer examples for choice-level analysis; removed commented-out code.
- Improved description in DESCRIPTION to highlight both profile-level and choice-level estimators, IRR correction, and visualization.
Updated README and pkgdown site to emphasize choice-level analysis as the default and recommended framework.

Fixed reshape_projoint() docs (\textrightarrow macro removed).
Standardized return values for summary(), plot(), and print() methods to match CRAN policies.

Fixed DESCRIPTION metadata mismatch between Author and Authors@R fields, as requested by CRAN.

DESCRIPTION: aligned Author: exactly with CRAN’s derived Authors@R format (names and ORCID link formatting) to remove NOTE.

Robust SEs & clustering: fixes to variance options; users can now pass arguments supported by estimatr::lm_robust() for clustering and standard errors.
reshape_projoint() overhaul: now robust to arbitrary base-task orders, infers the repeated base task from the first base outcome, and requires the repeated outcome to be the last element of .outcomes.

Minor bug fixes and documentation improvements.
Major redesign of the website, including improved vignettes and a clearer step-by-step tutorial structure.
Fixed input validation for specifying inter-rater reliability (IRR), ensuring that IRR values must be between 0.5 and 1.