Facet reliability and separation snapshot bar plot

Compact facet-level visual of the Wright & Masters (1982) separation, strata, and reliability indices that diagnose_mfrm() computes. Helpful as a single small figure for "are persons / raters / criteria distinguishable?" review. These are Rasch/FACETS-style separation indices on the fitted logit scale, not ICCs; use compute_facet_icc() for the complementary observed-score variance-share view.

Usage

plot_reliability_snapshot(
  fit,
  diagnostics = NULL,
  metric = c("reliability", "separation", "strata"),
  preset = c("standard", "publication", "compact", "monochrome"),
  draw = TRUE
)

Arguments

fit: An mfrm_fit from fit_mfrm().
diagnostics: Optional diagnose_mfrm() output. Computed on demand when omitted.
metric: "reliability" (default), "separation", or "strata".
preset: Visual preset.
draw: If TRUE, print the ggplot object.

Value

Invisibly, a ggplot object. plot_data(p, "table") is a tidy Facet, Metric, Value data frame. When present in the supplied diagnostics, precision tier, reliability use/basis, formal-inference flag and model-summary notes are retained alongside each plotted value.

Details

High Person separation reliability indicates distinguishable abilities. High rater separation reliability indicates distinguishable severities, not agreement between raters. If raters should be interchangeable, reliable severity differences merit review. Low rater separation reliability can also reflect imprecise estimates and does not establish agreement. The index does not test every rater pair or measure the practical size of severity differences. Examine estimated severities and their uncertainty, and use interrater_agreement_table() for observed agreement.

The separation-reliability formula is shared across facets, but the facet's role changes its interpretation. Person separation reliability is analogous to test reliability, not numerically identical to Cronbach's alpha; MML Person rows using EAP/posterior SD also have the limitations described in mfrmr-package. No universal adequate/strong reference lines are drawn. Reliability, separation and Strata use approximate marginal-SE variance subtraction and do not propagate covariance among estimates, even when the SEs come from joint information. Model and fit-adjusted indices are not confidence limits; Strata does not identify actual groups. Ordinary JML uses exploratory location precision; these indices are screening summaries, not structural confidence intervals. Corrected-JML RootSE is not an input to these indices. Corrected JML and two-family GPCM are unavailable here, including when diagnostics are supplied. The returned plot retains interpretation notes for saved reuse.

Session plot defaults

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

Migration in 0.2.4

This renderer is internal. Use the corresponding plot() route in mfrmr_plot_migration; the former exported name is no longer available.

References

Linacre, J. M. Facets documentation, Reliability - separation - strata. https://www.winsteps.com/facetman/reliability.htm

Examples

# Supplementary workflow: run explicitly to keep routine checks brief.
# example("plot_reliability_snapshot", run.dontrun = TRUE)
if (FALSE) { # \dontrun{
# Saved synthetic JML/RSM analysis of example_core.
# For your own data, fit a model first; see ?fit_mfrm.
example_fit <- readRDS(system.file("examples", "core-jml.rds", package = "mfrmr"))
fit <- example_fit$fit
p <- plot(fit, type = "reliability", draw = FALSE)
plot_data(p, "table")
# High Person value: distinguishable abilities.
# High rater value: distinguishable severities, not agreement.
# Low rater value: not proof of agreement; inspect estimation precision.
plot_data(p, "notes")
} # }