Review and plot portable fixed-calibration scores

These methods provide the first review surface for the result returned by score_mfrm_calibration(). print() gives a compact batch disposition, summary() exposes readable score and review tables, and plot() shows conditional score uncertainty or numerical review quantities without refitting a model.

Usage

# S3 method for class 'mfrm_calibration_score'
summary(object, digits = 3L, ...)

# S3 method for class 'mfrm_calibration_score'
print(x, ...)

# S3 method for class 'summary.mfrm_calibration_score'
print(x, ...)

# S3 method for class 'mfrm_calibration_score'
plot(
  x,
  type = c("interval", "precision", "edge_mass"),
  top_n = 40L,
  sort_by = c("estimate", "sd", "person"),
  label_review = TRUE,
  main = NULL,
  draw = TRUE,
  preset = c("standard", "publication", "compact", "monochrome"),
  ...
)

Arguments

object, x: An mfrm_calibration_score returned by score_mfrm_calibration().
digits: Number of decimal places in the summary.
...: Reserved for generic compatibility.
type: One of "interval", "precision", or "edge_mass".
top_n: Maximum number of scored Persons shown. Review rows are selected first when truncation is necessary. Use Inf to show all.
sort_by: Selection priority after review rows: absolute estimate, posterior SD, or Person identifier.
label_review: Whether review points are labelled in the precision and edge-mass views. Interval plots always label the displayed Persons.
main: Optional plot title.
draw: If TRUE, print the ggplot figure on the active graphics device. The ggplot object is returned invisibly in either case.
preset: Visual preset: "standard", "publication", "compact", or "monochrome".

Value

print() returns its input invisibly. summary() returns a summary.mfrm_calibration_score containing overview, estimates, review, row_review, settings, and notes. When adaptive integration was requested and scored rows exist, it also retains unrounded quadrature_review and a compact quadrature_overview. When cut scores were requested, decisions preserves every probability, cut boundary and actual scoring prior without rounding; unavailable Persons remain in that table with NA probabilities. plot() returns a ggplot object (invisibly) whose underlying payload (accessible via plot_data()) retains the selected plotting table, selection accounting, unplotted Person dispositions, interpretation guidance, and plotting settings.

Details

The default interval plot shows posterior EAP estimates and central intervals. type = "precision" plots the number of valid response rows against posterior SD. type = "edge_mass" compares posterior mass on the two outer quadrature nodes with the recorded review threshold. Review dispositions are highlighted in every view.

These are score-batch review displays, not calibration-fit diagnostics. Posterior SDs and intervals are conditional on the frozen point calibration and the actual scoring prior. They exclude calibration-parameter uncertainty. For GPCM or an explicitly supplied scoring prior, summaries and plot data retain both original and actual prior values without rounding, the per-Person integration checks, and (for GPCM) the source decision recorded at extraction. These records describe conditional scoring, not a new evaluation of the training fit or the suitability of its population for another cohort. summary(x)$settings$score_integration_review compares the reported EAP/SD with higher-order adaptive references. The separate quadrature_review compares adaptive and fixed grids; for an adaptive scorer, that fixed grid did not produce the reported scores. Printing distinguishes the two checks. Persons with no valid responses have no score coordinate and are retained in summary(x)$review and in the plot payload's unplotted_dispositions component.

The summary object retains every returned score and review disposition. Its estimates table preserves the available estimate and uncertainty basis columns, calibration identifiers, scoring algorithm and requested interval level when extracted or written to CSV. The interval level is not rounded. Older score results recover these fields from their recorded settings when re-summarized. Saved grid-based intervals retain their original endpoints; printed results and interval plots explain that their posterior mass can differ from the requested level. Its print method shows at most ten rows from each table so routine console output stays compact. Plots distinguish scored and review states by shape as well as colour. plot(x) returns a ggplot object directly; as_ggplot(p) and as_ggplot(p, component = "data") preserve the same selected view, saved bounds, review symbols and unplotted dispositions for both ggplot and legacy mfrm_plot_data objects. Use plot_data() to extract other components. Changing the interval level requires an explicit scoring call; conversion does not recompute intervals. Notes remain in the returned object and are not automatically drawn.

Session plot defaults

See mfrmr_visual_diagnostics for options(mfrmr.plot_preset), argument precedence, and appearance when saving or reusing a plot.

Examples

# Long-running workflow; run explicitly when needed.
# example("mfrm_calibration_score_methods", run.dontrun = TRUE)
if (FALSE) { # \dontrun{
dat <- load_mfrmr_data("example_core")
ids <- unique(dat$Person)
training <- dat[dat$Person %in% ids[1:18], , drop = FALSE]
fit <- fit_mfrm(
  training, "Person", c("Rater", "Criterion"), "Score",
  model = "RSM", method = "MML", quad_points = 31
)
q_review <- mml_quadrature_sensitivity(
  fit, training, quad_points = c(31, 61), theta_points = 41
)
summary(q_review) # Inspect movement; these grids are not universal cutoffs.
fit <- q_review$fits$q61
calibration <- update(status = "frozen", 
  update(status = "validated", extract_mfrm_calibration(
    fit, quadrature_review = q_review, scoring_quad_points = 121
  ))
)
new_rows <- dat[dat$Person %in% ids[19:20], , drop = FALSE]
scores <- score_mfrm_calibration(calibration, new_rows,
  cut_scores = c(Basic = -0.5, Advanced = 0.8))
summary(scores)
apa_table(scores, which = "decisions")
build_summary_table_bundle(scores)$table_index
plot(scores, type = "interval", draw = FALSE)
plot(scores, type = "edge_mass", draw = FALSE)
} # }