Skip to content

NEWS.md

Latest commit

 

History

History
74 lines (55 loc) · 6.03 KB

NEWS.md

File metadata and controls

74 lines (55 loc) · 6.03 KB
Raw

jaspReliability Changelog

HOW TO READ AND UPDATE THIS CHANGELOG:

This document follows a modified Keep a Changelog format adapted for the R/JASP ecosystem. Releases are listed in reverse chronological order (newest first). As an example see jaspModuleTemplate

  • Adding New Changes (For Contributors): All new commits should be logged at the very top of the file under the # jaspModuleTemplate (development version) header. Place your bullet point under the appropriate category (## Added, ## Fixed, etc.).
  • Issue References: Please reference the relevant GitHub Issue (if any) at the end of your line (e.g., ([Issue #19](https://github.com/jasp-stats/jaspModuleTemplate/issues/19)).
  • Format Categories: > * Added: New template features, QML examples, or build tools.
    • Changed: Updates to default configurations, boilerplate code, or dependencies.
    • Fixed: Bug fixes in the build pipeline, R wrappers, or QML layouts.
    • Deprecated / Removed: Outdated template components or legacy code.

jaspReliability (development version)

Added

  • Added Bayesian Multidimensional Reliability analysis: multidimensional omega coefficients via Bayesrel::bomegas, with scale/item tables, model fit, posterior, trace, and posterior predictive check plots (Issue #1336). The item table reports a point estimate and credible interval for both if-item-dropped omegas, and the prior on the item and structural loadings has a variance scale in addition to a mean. Requires Bayesrel >= 0.7.9.
  • Dispersal prior now wired through Bayesrel prior samplers (replaces deprecated fitCutoffNull).

Changed

  • Aligned multidimensional option names with unidimensional counterparts for consistency.
  • Renamed unidimensional reliability analyses: unidimensionalReliabilityFrequentist -> reliabilityUnidimensionalFrequentist, unidimensionalReliabilityBayesian -> reliabilityUnidimensionalBayesian (upgrade mappings added).

Fixed

  • Fixed bi-factor model crash caused by param.out = TRUE in Bayesrel::bomegas.
  • Fixed missing dependOn declarations for point-estimate option on scale table and latentCorDf in base container.
  • Fixed untranslated point-estimate column title in scale/item tables.
  • Rater Agreement: declared factor level order is now authoritative for weighted Cohen's kappa and ordinal Krippendorff's alpha even when the labels look numeric (e.g. levels "1","3","2" declaring low < medium < high); numeric sorting is only used for columns with no declared order. Contradictory numeric-looking declared orders across raters now show the common-scale error instead of being silently sorted.
  • Rater Agreement: pairwise weighted Cohen's kappa now uses the full declared common scale for every rater pair, so a pair that never observes an interior category still gets that category's correct distance/weight instead of a rescaled subset.
  • Rater Agreement: Kendall's W (table and bootstrap CI) now rejects raters-in-rows data whose merged ordinal scale is ambiguous, matching the existing policy for weighted Cohen's kappa and ordinal Krippendorff's alpha; previously the result depended on the order of the subject columns.

jaspReliability 0.97.1

Added

  • Rater Agreement: added Kendall's W coefficient with bootstrap CI support.
  • Rater Agreement: added the F test for Kendall's W alongside the chi-square test (Issue #2151).

Fixed

  • Rater Agreement: declared ordinal level order is now respected (Kendall's W, Krippendorff's alpha, weighted Cohen's kappa), including with raters in rows and when raters use different subsets of the categories; validation runs on the analyzed (complete/pairwise) data; degenerate inputs (all-missing, constant ratings, single category, one rater) show clear errors instead of failing or reporting invalid results.
  • Rater Agreement: Krippendorff's alpha is validated on the ratings that actually enter a coincidence (subjects rated at least twice); previously such data could report a non-estimable coefficient together with a bootstrap confidence interval.
  • Rater Agreement: Fleiss' kappa standard errors are computed from the Fleiss, Nee & Landis (1979) formulas instead of being reconstructed from rounded output.

Changed

  • Rater Agreement: grouped bootstrap samples, CI level, and seed into an "Advanced Options" section; removed pre-checked defaults for all coefficients; added placeholder table prompting users to select a coefficient when variables are assigned.
  • Rater Agreement: the tie correction for Kendall's W is enabled by default, and bootstrap confidence intervals are only available for the tie-corrected coefficient, because resampling with replacement introduces ties.
  • Rater Agreement: weighted Cohen's kappa and ordinal Krippendorff's alpha require the raters to share one ordinal scale. When the declared category orders are contradictory or do not determine a unique order, an error is shown instead of an order-dependent result.
  • Rater Agreement: Cohen's kappa validates and reports each rater pair on its own pairwise-complete data. Pairs with fewer than three jointly rated subjects, without variation, or with a non-estimable coefficient are excluded from the average kappa and listed in a note. Pair rows are ordered V1-V2, V1-V3, V1-V4, ...

jaspModuleTemplate 0.2.0

Added

  • Added NEWS.md
  • Added workflow to remind users to update their NEWS.md.
  • Added workflow to auto-bump version when user does not do so.

jaspModuleTemplate 0.1.0

Added

  • Initial examples to showcase JASP module development

Changed

  • Use best practices for checking input (Issue #19).
  • The main results table now defaults to displaying 95% Confidence Intervals for effect sizes.

Fixed

  • Remove deprecated dependencies from qml files (Issue #14).