NEWS

integrity 1.0.3 (2026-10-20)

Changed item 2.1 to a two-sided Wald-Wolfowitz runs test, aligning it with the published IPD Integrity Tool. Small or unbalanced category counts use exact conditional p-values; larger counts use a normal approximation.
Item 2.1 detailed output now reports observed and expected runs, missing counts, p-values, and test methods instead of adjacent-pair counts. This changes the columns in detail_tables[["2.1"]].
Item 2.1 now sorts by the nominated randomisation date, retaining row order for ties. It warns and falls back to row order if no usable dates are available. Partly missing dates are excluded from this item with a warning.
Clarified omission of missing baseline values and interpretation of untestable variables for item 2.1; recorded the ordering and date exclusions in the detailed output.

Added baseline-pattern plots for item 1.1, with a caution when no usable randomisation date is supplied.
Gave terminal-digit plots independent y-axis scales for each variable.
Changed allocation plots to show cumulative participant counts.
Preserved continuous types for numeric baseline imbalance checks and reported untestable binary baseline checks as skipped.
Fixed handling of all-missing columns and single-participant rare outcomes.
Declared the cardx dependency needed for gtsummary statistical tests.
Updated the vignette with download and adaptation instructions and loading through the installed package.
Corrected Anna Lene Seidler's name and updated author ordering.
Added regression checks and excluded saved vignette workspaces from builds.