NEWS

rbmiUtils 0.3.0.9000

describe_draws() extracts structured metadata from rbmi draws objects, including method type, formula, sample count, and (for Bayesian methods) MCMC convergence diagnostics (ESS, Rhat).
describe_imputation() extracts imputation metadata including method, number of imputations (M), reference arm mappings, and a missingness breakdown by visit and treatment arm.
pool_to_ard() gains an analysis_obj parameter that enriches the ARD with MI diagnostic statistics (FMI, lambda, RIV, Barnard-Rubin adjusted df, relative efficiency) when the pooling method is Rubin's rules.

efficacy_table() gains font_family, font_size, and row_padding parameters for publication-ready table styling.
plot_forest() gains font_family and panel_widths parameters for customizable typography and panel layout.
plot_forest() left panel now uses left-aligned text (hjust=0) for consistent positioning regardless of label length.
Non-Rubin pooling methods now emit an informative cli::cli_inform() message when MI diagnostics are not applicable, rather than returning NA rows.

End-to-end pipeline vignette "From rbmi Analysis to Regulatory Tables" with continuous and binary analysis walkthroughs.
README enhanced with rendered efficacy table and forest plot visual teasers.
Rendered example output images in plot_forest() and efficacy_table() help pages.
Inline cross-references to rbmi and beeca documentation in all vignettes.

validate_data() now uses cli-formatted error messages with clearer guidance for malformed interaction terms, empty data, and type mismatches.
prepare_data_ice() now errors immediately when vars$strategy is NULL instead of silently using a default column name.
prepare_data_ice() warns when visit column is character with guidance to convert to factor for correct ordering.
validate_data() batches all type coercion warnings into a single message and warns on all-NA covariate columns.
Data preparation functions handle edge cases (single subject, single visit, all-NA outcome, all-complete data) gracefully.

tidy_pool_obj() now uses regex-based parameter parsing instead of splitting on _. Output columns (parameter_type, lsm_type, visit) now contain the full visit name rather than truncated fragments.

efficacy_table() creates regulatory-style gt summary tables from pool objects.
plot_forest() creates publication-quality three-panel forest plots from pool objects.
pool_to_ard() converts pool objects to pharmaverse ARD format via the cards package.
print() and summary() S3 methods for pool and analysis objects.
create_impid() converts lists of imputed datasets into stacked data.frames.
combine_results() combines tidy results from multiple analyses.
format_results() and format_results_table() for publication-ready formatting.
extract_trt_effects() and extract_lsm() convenience filters for tidy results.

Added cli (>= 3.6.0) and lifecycle (>= 1.0.4) to Imports for improved error messaging and deprecation support.

Enhanced input validation in analyse_mi_data() with clearer error messages using inherits() instead of internal class helpers.
Enhanced input validation in prepare_data_ice() to check vars fields.
Cross-references between related functions in documentation.
Expanded test coverage and integration tests.

reduce_imputed_data() and expand_imputed_data() for efficient storage of imputed datasets.
validate_data() for pre-flight validation before rbmi::draws().
prepare_data_ice() to build data_ice from flagged ICE columns.
summarise_missingness() for tabulating missing data patterns.
format_pvalue(), format_estimate(), format_results_table() formatting utilities.
ADEFF and ADMI example datasets.