Provenance columns in posterior tables

Posterior tables retain the existing estimate and identifier columns and add ordinary columns describing how the result was computed. These columns are included in default reports and CSV exports; compact reports omit most of them. Unknown or inapplicable values are typed NA.

Columns

• analysis_type: response, hypothesis, joint hypothesis, simultaneous band, representative selection, response timing, or diagnostic analysis.

• interval_probability and interval_method: the mass and construction of intervals actually reported, equal_tailed or empirical_sup_norm.

• scale_by, scale_var, and scale_factor: response normalisation. A shock_sd response is divided by the reported sample standard deviation.

• units: known units such as FEVD percentages or response-timing periods; economic units are left unknown when they cannot be recovered.

• source_package and source_version: the upstream package used for the current computation. They do not reconstruct estimation software history.

• n_draws: number of posterior draws used for the result.

• draw_id: original draw labels in draw-level tables. The integer draw column retains positional indexing.

• selected_draw_id: original label of a representative draw; draw_index remains its positional index.

Existing object_type, draws, and compare attributes remain supported. Legacy tables without provenance columns can still be plotted when their analysis is unambiguous from the existing columns and attributes.

See also

tidy_irf(), report_table(), write_bsvar_csv()