Build a model-choice review across RSM, PCM, and GPCM fits

Usage

build_model_choice_review(
  ...,
  labels = NULL,
  run_weighting_review = NULL,
  theta_range = c(-6, 6),
  theta_points = 61L,
  top_n = 10L,
  warn_constraints = TRUE
)

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

Arguments

...

For build_model_choice_review(): Two or more fitted mfrm_fit objects from fit_mfrm().

For summary(): Reserved for generic compatibility.

labels

Optional labels for the supplied fits. If omitted, names from ... are used when available; otherwise labels are generated from model/method combinations.

run_weighting_review

Logical. If TRUE and the supplied fits include at least one RSM/PCM reference plus one GPCM fit, also run build_weighting_review() for the first such pair.

theta_range, theta_points, top_n

Passed to build_weighting_review() when run_weighting_review = TRUE.

warn_constraints

Passed to compare_mfrm().

object

Output from build_model_choice_review().

digits

Number of digits for printed numeric values.

Value

An object of class mfrm_model_choice_review.

For summary(): An object of class summary.mfrm_model_choice_review.

Details

build_model_choice_review() is a user-facing synthesis helper. It does not estimate new models. It bundles:

compare_mfrm() for verified AIC/Person-BIC/SABIC/log-likelihood comparison;
comparison-boundary warnings captured from compare_mfrm() and retained in comparison_warnings for printing and appendix export;
model-role guidance for RSM, PCM, and GPCM;
reported step/slope coordinate counts, identified free-parameter counts, and the stored fit-readiness decision for every supplied model;
downstream-route availability for APA output, score-side export, linking, recovery, fair averages, bias screening, and summary-appendix handoff;
report wording templates that avoid treating better GPCM fit as an automatic operational-scoring decision;
mfrmr_output_guide("gpcm_capabilities") when GPCM is present;
optionally, build_weighting_review() for the first Rasch-family reference versus GPCM pair.

The word "bounded" describes the documented model and workflow scope: the package does not implement every possible generalized partial-credit many-facet extension. It does not mean that finite optimizer box bounds define the estimator. The current route uses positive slopes, allows separate slope/step owners in MML (JML requires slope_facet == step_facet), identifies slopes on the log scale with geometric mean 1, and keeps several downstream score-side/reporting helpers outside the documented boundary. Model-choice ranking also requires the current selectable IC contract: q<31 fits retain raw criteria but produce a screening/review-only bundle.

summary() method: Summarize a model-choice review

Examples

toy <- load_mfrmr_data("example_core")
fit_rsm <- fit_mfrm(toy, "Person", c("Rater", "Criterion"), "Score",
                    method = "MML", model = "RSM", quad_points = 31)
fit_pcm <- fit_mfrm(toy, "Person", c("Rater", "Criterion"), "Score",
                    method = "MML", model = "PCM", step_facet = "Criterion",
                    quad_points = 31)
review <- build_model_choice_review(RSM = fit_rsm, PCM = fit_pcm)
summary(review)
#> mfrm Model Choice Review
#> 
#> Overview
#>  FitCount   Models OperationalReference SensitivityModel
#>         2 RSM, PCM                  RSM             <NA>
#> 
#> Next Actions
#>  - Read comparison_table for fit evidence, but decide operational use from the
#>    score interpretation.
#>  - Use downstream_routes before calling APA, score-side export, linking,
#>    recovery, fair-average, or bias-screening helpers.
#>  - Use report_templates when drafting methods text so the fitted model and
#>    score contract are described accurately.
#> 
#> Comparison Table
#>  Label Model Method Persons Npar   LogLik      AIC      BIC    SABIC Delta_AIC
#>    RSM   RSM    MML      48    8 -900.013 1816.025 1830.995 1805.897     1.527
#>    PCM   PCM    MML      48   14 -893.249 1814.498 1840.695 1796.774     0.000
#>  Delta_BIC Delta_SABIC InferenceReady
#>        0.0       9.123           TRUE
#>        9.7       0.000           TRUE
#>   Full IC audit fields remain available in `$comparison_table`.
#> 
#> Model Contracts and Readiness
#> 
#> Step parameters (reported values versus independent parameters)
#>  Label Model          StepStructure Reported Free
#>    RSM   RSM          shared ladder        3    2
#>    PCM   PCM facet-specific ladders       12    8
#> 
#> Slope parameters (reported values versus independent parameters)
#>  Label Model SlopeStructure Reported Free
#>    RSM   RSM     fixed at 1        0    0
#>    PCM   PCM     fixed at 1        0    0
#> 
#> Fit readiness
#>   RSM: numerical convergence: Passed.
#>   PCM: numerical convergence: Passed.
#>  Label Model FormalInference
#>    RSM   RSM              No
#>    PCM   PCM              No
#>                                                          Interpretation
#>  Fit-readiness requirements satisfied; formal precision review required
#>  Fit-readiness requirements satisfied; formal precision review required
#>   Full score contracts and readiness reasons remain in `$model_roles`.
#> 
#> Downstream Routes
#>  Label Model FullAPARoute ScoreSideExport LinkingSynthesis RecoveryChecks
#>    RSM   RSM    Available       Available        Available      Available
#>    PCM   PCM    Available       Available        Available      Available
#>  FairAverage BiasScreening SummaryAppendix
#>    Available     Available       Available
#>    Available     Available       Available
#> 
#> Weighting Review Status
#>  Requested Available
#>      FALSE     FALSE
#>                                                                                   Message
#>  Not requested; set `run_weighting_review = TRUE` for the first RSM/PCM versus GPCM pair.
#> 
#> Notes
#>  - This review is a decision aid; it does not refit models or choose an
#>    operational model automatically.
#>  - Observation weights and GPCM discrimination-based reweighting are separate
#>    concepts.
#>  - Use GPCM wording only for the current constrained implementation, not for an
#>    unrestricted GPCM family claim.