Changelog • qgcomp.multi

qgcomp.multi 0.5.0

Added estimand-scale choices for ordinary and multiple-imputation fits, including odds-ratio and rate-ratio summaries for supported family/link combinations.
Added percentile and basic bootstrap confidence intervals for single-fit coefficient summaries and MSM-based prediction intervals.
Added confregion() for single-fit bootstrap-covariance chi-squared confidence regions for selected MSM coefficients.
Added an effect-scales vignette explaining fitting scale, display scale, response-scale prediction, interval methods, and current MI limits.

Updated binomial-logit and Poisson-log defaults to use ratio estimands unless the user requests additive alternatives.
Clarified reporting so ratio estimands show both the fitting-scale coefficient and the exponentiated ratio where appropriate.
Kept prediction and surface plotting on the response scale by default while allowing direct MSM contrasts on the estimand scale.

Expanded function documentation, README, workflow material, and pkgdown configuration for the v0.5.0 inference features.

Added native multiple-imputation fitting and pooling through qgcomp.glm.multi.mi().
Added support for multiply imputed inputs from completed-data lists and mice::mids objects.
Added broom-style tidy() and glance() methods for single-fit and pooled MI objects.
Added optional bootstrap-level parallel execution for ordinary fits and native multiple-imputation workflows.

Expanded the public workflow beyond single-fit estimation to include pooled MI inference and runtime-oriented parallel controls.
Formalized the pooled MI result structure around a dedicated fitted-object class for summary and inference tasks.

Expanded the README and package documentation to cover native multiple-imputation workflows, parallel execution, and broom-based reporting.

Added MSM-based prediction support, including direct contrasts and exact fit-time surface extraction.
Added surface plotting for fitted intervention-response surfaces.
Added intervention support, bootstrap, and adequacy diagnostics as first-class public features.
Added sensitivity helpers for Monte Carlo size, bootstrap size, and quantization choice.
Added the main applied workflow vignette for end-to-end package use.

Extended fitted qgcompmulti objects to retain fit-time prediction information needed for prediction, plotting, and diagnostic methods.
Reorganized the package around a clearer post-fit workflow built on stored intervention grids and surfaces.

Expanded examples and method documentation for prediction, plotting, diagnostics, and sensitivity analysis.

Introduced the structured qgcompmulti fitted-object class.
Added core fitted-model methods including print(), summary(), coef(), vcov(), confint(), formula(), and nobs().
Added a cleaner object contract to support future extensions in prediction, diagnostics, and multiple-imputation workflows.

Moved the package away from an earlier proof-of-concept output structure toward a more conventional model-object interface.
Standardized the organization of fitted-model results and metadata.

Updated the package documentation to reflect the new fitted-object structure and model-method workflow.