Portable fixed-calibration capabilities
Source:R/api-calibration.R
mfrm_calibration_capabilities.RdReturns the model and estimator combinations supported by the portable
calibration workflow. This matrix concerns saved calibration artifacts;
it does not replace the wider fitted-object capabilities of fit_mfrm() or
predict_mfrm_units().
For JML fits, fitted-object scoring is post-hoc EAP with a standard-normal
reference prior by default, or an explicit scoring_prior. It is not
ML/WLE scoring or a population distribution estimated by JML. Portable
extraction supports RSM/PCM and shared-owner GPCM JML within their distinct
source-check scopes. No artifact stores training Person estimates.
Value
A data frame with one row per model, estimator, and scoring-basis
combination. PortableCalibration is either "available" or
"unavailable"; the listed scope and source checks still apply.
Examples
mfrm_calibration_capabilities()
#> Model Estimator ScoringBasis
#> 1 RSM MML fixed standard normal
#> 2 PCM MML fixed standard normal
#> 3 RSM/PCM MML estimated population or latent regression
#> 4 GPCM MML frozen estimated intercept-only normal
#> 5 RSM/PCM JML post-hoc standard normal reference
#> 6 GPCM JML post-hoc standard normal reference
#> 7 GPCM Corrected JML post-hoc standard normal reference
#> 8 GPCM (two families) MML fixed standard normal
#> PortableCalibration AnchorSupport
#> 1 available stored direct and group facet anchors
#> 2 available stored direct and group facet anchors
#> 3 unavailable not available for portable calibration
#> 4 available not supported
#> 5 available not supported
#> 6 available not supported
#> 7 available not supported
#> 8 available not supported
#> InteractionSupport
#> 1 stored two-way facet interactions
#> 2 stored two-way facet interactions
#> 3 not available for portable calibration
#> 4 not supported
#> 5 not supported
#> 6 not supported
#> 7 not supported
#> 8 not supported
#> ExistingAlternative
#> 1 portable artifact or fitted-object scoring
#> 2 portable artifact or fitted-object scoring
#> 3 use fitted-object scoring with the fitted population model
#> 4 conditional portable artifact or fitted-object GPCM scoring
#> 5 portable artifact or fitted-object post-hoc EAP; not ML/WLE
#> 6 conditional portable artifact or fitted-object post-hoc EAP; not ML/WLE
#> 7 experimental corrected calibration with post-hoc EAP; not corrected Person ML/WLE
#> 8 experimental two-family conditional artifact or fitted-object EAP
#> Limitation
#> 1 one observed score scale, one latent dimension, known facet levels, and an explicit same-data quadrature review
#> 2 one observed score scale, one latent dimension, known facet levels, and an explicit same-data quadrature review
#> 3 population coding and conditional parameters are not stored in the artifact
#> 4 passing conditional source checks; known levels, unit weights, no anchors, interactions or latent regression
#> 5 finite identified RSM/PCM JML source; unit weights, no anchors or interactions; reference prior is not estimated by JML
#> 6 shared owners, unit weights, no anchors/interactions; passing local JML checks; incomplete global audits remain recorded
#> 7 shared owners; explicit correction order; passing adjusted-equation/root checks; residual calibration bias may remain; no calibration uncertainty propagated
#> 8 two ordered slope owners, second-owner steps, fixed N(0,1), known levels, unit weights and passing source/batch checks; no prior override or repeated-event extension; calibration uncertainty and population transport are not qualified