Display saved GPCM intervals and bootstrap results — plot.mfrm_slope

Plot the selected inference result without fitting or recomputing covariance. These plots describe discrimination or specified contrasts, not rater severity, agreement, quality or recommended scoring weights.

Usage

# S3 method for class 'mfrm_slope_intervals'
plot(
  x,
  title = "GPCM slope uncertainty",
  subtitle = paste(attr(x, "target"), attr(x, "method"), paste0(100 * attr(x, "level"),
    "%"), if (attr(x, "simultaneous") == "none") "pointwise" else "Bonferroni", sep =
    " | "),
  reference = NULL,
  draw = TRUE,
  ...,
  type = c("interval", "profile"),
  caption = NULL
)

# S3 method for class 'mfrm_gpcm_bootstrap'
plot(
  x,
  title = "GPCM bootstrap inference",
  subtitle = "Fitted-model parametric bootstrap; unresolved refits retained",
  draw = TRUE,
  ...
)

Arguments

x: Saved confint.mfrm_fit() output from parm = "slopes", or bootstrap_mfrm_gpcm() output.
title, subtitle: Optional returned metadata; NULL removes it. The interval description defaults to the saved target, method, confidence level and any saved numerical or bootstrap cautions. Custom text overrides this default, while the diagnostics remain in the plot data and interval tables. Titles and notes are not drawn. Use plot_data(p, "notes") for descriptions and cautions, and plot_data(p, "title") for the suggested figure title.
reference: Optional finite numeric vertical reference. Default NULL; no quality threshold or scoring-policy cutoff is selected.
draw: TRUE draws the ggplot; FALSE returns it only.
...: For bootstrap slope results, passed to confint() to select its target and level. Otherwise unused.
type: For saved slope intervals, "interval" displays bounds; "profile" displays saved likelihood evaluations and the same-target Wald limits. The default is "profile" for saved profile calculations and "interval" otherwise. This never initiates profiling.
caption: For saved slope intervals, optional returned caption. NULL removes it. Two-family intervals state the distinct family references by default. Profile plots supply a default explanation of line styles and missing coordinates when omitted.

Value

Invisibly a ggplot, retaining an mfrm_plot_data attribute for plot_data(). No fitting or resampling is performed.

Details

Crosses retain unavailable intervals. Open circles retain intervals with infinite endpoints; the plot does not replace them with finite bounds. Incomplete intervals retain the computed endpoint as a tick, without a connecting interval line. A missing endpoint is not an infinite endpoint or a calibrated one-sided interval. Finite review-only bounds use a dashed line. The table retains IntervalStatus, CIEligible, CIUse (when saved), all bounds and failure reasons; returned notes explain the display. For a bootstrap LRT, a histogram shows resolved null statistics and a line the observed statistic. The returned notes count every unresolved replicate.

Use apa_table() for target-aware tables, plot_data() for plotted values, and as_ggplot() for customization. Attach one result or a named list with mfrm_results(fit, intervals = list(slopes = ci, curves = curves), compute = "never"). Named plot routes become gpcm_slopes, gpcm_curves, etc. Exact source matching is required; default fit diagnostics are preserved. export_mfrm_results() saves these tables and plots; replay reloads the saved RDS, preserving the selected methods without rerunning estimation. Markdown report summaries include the finite, unbounded and unavailable interval counts and the saved inference cautions/reasons.

For positive slope or ratio intervals spanning many orders of magnitude, use as_ggplot(ci) + ggplot2::scale_x_log10(). Finite bounds and estimates must be strictly positive. Keep a linear axis for signed differences. Changing the axis does not narrow the interval or make it more reliable.

Examples

# ci <- confint(fit, parm = "slopes", scale = "standardized", method = "sandwich")
# plot(ci, title = NULL)
# apa_table(ci)
# plot_data(ci, "table")