mfrmr cheatsheet (2 pages)

In mfrmr: Estimation and Diagnostics for Many-Facet Measurement Models

knitr::opts_chunk$set(
  eval = FALSE, echo = TRUE, message = FALSE, warning = FALSE,
  collapse = TRUE, comment = "#>"
)

Quick reference: mfrmr 0.2.0

A two-page reference for the fit / diagnose / report loop in mfrmr. Every snippet assumes library(mfrmr) and a long-format data.frame with Person, one or more facet columns, and Score.

1. Fit

toy <- load_mfrmr_data("example_core")

fit <- fit_mfrm(
  toy,
  person = "Person",
  facets = c("Rater", "Criterion"),
  score  = "Score",
  method = "MML",          # default; "JML" for exploratory speed
  model  = "RSM"            # "PCM", "GPCM" (bounded) also supported
  # quad_points = 31 (default, publication tier; 7/15 for iteration)
)

2. Diagnose

diag <- diagnose_mfrm(fit, residual_pca = "none")
# diagnostic_mode = "both" (default) runs residual + strict marginal.
summary(fit)
summary(diag)

3. Report / APA

chk  <- reporting_checklist(fit, diagnostics = diag)
apa  <- build_apa_outputs(fit, diag)
cat(apa$report_text)

tbl  <- apa_table(fit, which = "facets", diagnostics = diag)
kab  <- as_kable(tbl)       # Markdown / HTML via kableExtra
ft   <- as_flextable(tbl)   # Word / PowerPoint via flextable

4. Bias / DFF / Equivalence

bias <- estimate_bias(fit, diag, facet_a = "Rater", facet_b = "Criterion")
summary(bias)
plot_bias_interaction(bias, plot = "ranked", show_ci = TRUE)

dff <- analyze_dff(fit, diagnostics = diag,
                   facet = "Rater", group = "Group", data = toy)
plot_dif_summary(dff)

eq <- analyze_facet_equivalence(fit, facet = "Rater", ci_level = 0.95)
plot_facet_equivalence(eq, type = "forest")

5. Hierarchical structure and small-N review

review <- facet_small_sample_review(fit)
summary(review)

icc <- compute_facet_icc(
  toy, facets = c("Rater", "Criterion"),
  score = "Score", person = "Person",
  ci_method = "profile"      # "boot" for parametric bootstrap CI
)

h <- analyze_hierarchical_structure(
  toy, facets = c("Rater", "Criterion"),
  ci_method = "profile"
)

6. Empirical-Bayes shrinkage for small-N facets

# Integrated:
fit_eb <- fit_mfrm(toy, "Person", c("Rater", "Criterion"), "Score",
                    method = "MML", facet_shrinkage = "empirical_bayes")
shrinkage_report(fit_eb)

# Post-hoc:
fit_eb2 <- apply_empirical_bayes_shrinkage(fit)

7. Plot surface (selected)

| Function | Purpose | |---|---| | plot(fit) | Wright map (default) | | plot(fit, type = "bundle") | Wright + pathway + CCC | | plot(fit, type = "ccc_overlay") | Model + observed theta-bin dots | | plot(fit, type = "wright", group=..., group_data=...) | DIF density overlay | | plot(fit, type = "shrinkage") | Shrinkage direction arrows | | plot_threshold_ladder(fit) | Category threshold ordering | | plot_person_fit(fit, diag) | Per-person Infit/Outfit bubble | | plot_rater_severity_profile(fit, diag) | Rater ranking + CI bands | | plot_bias_interaction(bias, plot="heatmap") | Bias cell heatmap | | plot_displacement(fit, show_ci=TRUE) | Anchor displacement lollipop | | plot_fair_average(fit, show_ci=TRUE) | Fair averages; CIs for Rasch-family fits | | plot_anchor_drift(drift, type="forest") | Wave-level anchor CI forest | | plot.mfrm_equating_chain(chain, type="graph") | Bipartite wave/anchor graph | | plot_apa_figure_one(fit) | Publication 2x2 composite | | plot_dif_summary(dff) | DIF effect-size summary |

8. Reproducibility and export

export_mfrm_bundle(fit, diagnostics = diag, path = "mfrmr_bundle/")
manifest <- build_mfrm_manifest(fit, diagnostics = diag)
# manifest$environment, $dependencies, $input_hash, $session_info,
# $hierarchical_review, $missing_recoding, $shrinkage_review
replay_script <- build_mfrm_replay_script(fit, path = "replay_mfrmr.R")

9. Missing-code pre-processing

fit <- fit_mfrm(
  dirty_data, "Person", c("Rater", "Criterion"), "Score",
  missing_codes = TRUE      # 99 / 999 / -1 / "N/A" / "" -> NA
)
fit$prep$missing_recoding

Package links

• Source and documentation: https://github.com/Ryuya-dot-com/mfrmr
• Citation: citation("mfrmr")

mfrmr 0.2.0 cheatsheet. Distributed under MIT License.

mfrmr documentation built on June 13, 2026, 1:07 a.m.