Changes in version 2.0.2 (2026-06-19)                  

  - Added \value documentation to the print(), summary() and plot()
    methods.
  - Replaced \dontrun{} with \donttest{} in examples, following CRAN
    feedback.

                        Changes in version 2.0.1                        

  - CRAN resubmission. Single-quoted software names and added
    parentheses to function names in the DESCRIPTION, and removed a
    broken LICENSE link from the README, following CRAN feedback. No
    changes to functionality.

                        Changes in version 2.0.0                        

Breaking changes

  - Minimum R version raised to 4.1.0.
  - prLogisticDelta() now takes a fitted model object as its first
    argument (instead of a formula + dataset). This aligns with standard
    R modelling conventions and allows the same function to handle all
    supported model types.
  - The cluster and pattern arguments of the old prLogisticDelta() are
    replaced by standardisation. Use standardisation = "conditional"
    (default) or standardisation = "marginal".
  - The Hmisc dependency has been removed (it was unused).
  - Return value is now a prLogistic S3 object rather than a plain
    matrix, with print, summary, coef, confint, and plot methods.

New features

  - Continuous covariates: reference value defaults to the sample median
    (configurable via ref_continuous = "mean"). The old implementation
    only supported binary (0/1) predictors.
  - Flexible baseline: use ref_values = list(var = value) to set the
    reference value of any predictor. This overrides both automatic
    defaults and the model's contrast coding.
  - GEE support (prLogisticGEE()): prevalence ratios for longitudinal /
    repeated-measures data fitted with geepack::geeglm(). Robust
    (sandwich) variance is used automatically.
  - Complex survey support (prLogisticSurvey()): prevalence ratios for
    data from complex survey designs fitted with survey::svyglm().
    Design-consistent variance is used automatically.
  - plot.prLogistic(): base-R forest plot for quick visualisation of PR
    estimates and confidence intervals.
  - Improved delta-method gradient for marginal standardisation: uses
    the full sample-averaged counterfactual gradient (not a single fixed
    row), giving more accurate variance estimates.

Bug fixes

  - Fixed incorrect variance calculation in pr_marginal() for models
    with more than one continuous covariate.
  - Negative confidence interval bounds are no longer silently truncated
    at 0 without a warning; a diagnostic warning is now issued.
  - pr.conditional() no longer crashes on glmerMod objects that use
    lme4::fixef() instead of @beta (the slot accessor was fragile across
    lme4 versions).

Internal changes

  - Complete rewrite using roxygen2 documentation.
  - All internal functions prefixed with . to avoid namespace pollution.
  - Comprehensive testthat (edition 3) test suite covering utilities,
    delta method, bootstrap, GEE, and survey model types.
  - pkgdown website with _pkgdown.yml configuration.
  - GitHub Actions CI/CD workflow for multi-platform R CMD check and
    automatic pkgdown deployment.

                  Changes in version 1.2 (2013-09-19)                   

  - Replaced lmer class with glmerMod for compatibility with lme4
    ≥ 1.0-4.

                  Changes in version 1.1 (2011-10-26)                   

  - Added NAMESPACE and ChangeLog files.
  - Replaced mean with colMeans() in pr.marginal().
  - Modified vcov() structure for S4 compatibility.

                  Changes in version 1.0 (2011-07-23)                   

  - Initial release.