NEWS

---

vacalibration 2.3.1

• cause_map() now tests input class with any(class(df) %in% ...) and does so before inspecting the number of columns. The previous check could raise "the condition has length > 1" for inputs carrying more than one class, such as a tibble of getIndivProb() output.
• Columns named cause or cause1 are standardised to cause. Output is unchanged for all algorithms and age groups.

vacalibration 2.3

Bug fixes

• Stan models updated to the array declaration syntax required from Stan 2.33 onward. Version 2.2 failed at model compile time for users with StanHeaders >= 2.33, which reached CRAN in September 2026, making the package unusable for them. Requires rstan (>= 2.26).
• Removed get_stan_seqcalib_q(), which referenced a model file that does not exist in the package.

Changes in behaviour

• With donotcalib_type = "learn" (the default), causes whose uncalibrated CSMF is below 1% or above 99% are now also excluded from calibration, in addition to causes whose misclassification rates vary by less than nocalib.threshold. Calibrated estimates may therefore differ from those produced by version 2.2.

Testing

• Added regression tests that parse the bundled Stan models, so a future Stan syntax change is caught by R CMD check rather than by users.

vacalibration 2.2 (2026-03-20)

• Separate plot_vacalib() function for plotting.
• Clean and informative progress output.
• Documentation moved into the .R sources.