NEWS

cofad 0.4.0 (2026-08-21)

Added participant-level r scores for within-subjects and mixed contrasts via within_score = "r". The existing L score remains the default: use L when response magnitude matters and r when agreement with the predicted pattern matters.
Corrected sums of squares, pooled error variance, and effect sizes for between-subjects designs with unequal group sizes. Raw and aggregated input now use the same formulas, and effect-size conversions preserve contrast direction.
Corrected the within-only r_contrast calculation to use the denominator degrees of freedom from its contrast test. The paper-ready report now uses the signed r_contrast consistently and relates its square to the contrast-specific eta-squared measures.
Extended competing-contrast support to the app. Favored and rival weights are standardized before their difference is analyzed, matching lambda_diff().
Added conservative detect_design() suggestions based on replication, crossing, and nesting. Ambiguous data deliberately fall back to manual model selection.
calc_contrast_aggregated() no longer requires an explicit data = NULL.
Added the corrected mixed-design data set rosenthal_tbl68_mixed; the historical rosenthal_tbl68 object remains available for compatibility.
Reduced hard dependencies by replacing small uses of dplyr, Hmisc, lifecycle, readr, rlang, stringr, and tibble with base R or foreign. The pipe remains available for backward compatibility.

Restored stable, editable table-based model and contrast inputs. The model table shows automatic design suggestions but always permits manual changes.
Added categorized between-subjects, within-subjects, and mixed examples. Each example loads the model roles and planned weights documented in its source; examples with published rival hypotheses open in competing mode.
Mixed designs now explicitly offer either a between × within contrast or a within contrast averaged across groups. Pure designs follow the roles in the model table without redundant activation controls.
Added detailed variance-decomposition F tables with ordinary and partial eta squared, calculation tooltips, directional paper-ready reports, and interactive Plotly partitions of variation. Mixed output is explicitly based on participants' derived within-contrast scores rather than raw repeated outcomes.
Reports can be copied as rich HTML with a plain-text fallback. F tables can be copied as aligned text or HTML and downloaded as dependency-free DOCX files.
Added APA 7 citation-copy controls for the tutorial and software paper. The package-level citation("cofad") command now returns both references.
Improved file validation, example-loading security, small-p-value formatting, table sizing and alignment, tooltips, collapsible panels, accessible colors, version display, and responsive copy controls.

Added the 2025 Behavior Research Methods tutorial as the primary citation, alongside the JOSS software paper.
Expanded the README with current R, app, Docker, and webR examples; equations now use GitHub's native mathematical notation.
Added a browser-only Shinylive/webR build with automatic GitHub Pages deployment and a containerized Shiny deployment through Dockerfile.
Expanded numerical, validation, design-detection, example-preset, citation, export, and in-process Shiny server tests. GitHub Actions now check multiple R versions and platforms and publish test coverage.
Verified the publication year and DOI metadata for Rosenthal, Rosnow, and Rubin (2000) and clarified the Sedlmeier and Renkewitz Table 16.1/16.2 distinction.

several bugs were fixed including rare occasions, where the order of factors was not treated correctly
included a helper function to calculate differences between two sets of lambdas for a competing contrast analysis, including documentation
added data examples
remove plyr as dependency
spell check
change maintainer to Johannes Titz, change order of authors, add contributors Mirka Henninger and Simone Malejka
improve summary functions, content and display
modify shiny GUI to use normal elements due to instability with moving elements (sortable)
migrate to shinytest2
include test for aggregated function, make it work with summary
deduplicate code (reuse between for mixed)
update summary for mixed

small improvements in documentation, references and paper for the official publication at journal of open source software