Draw observed-data MFRM resamples
Arguments
- spec
Output from
build_mfrm_resampling_spec().- keep_data
Logical; if
TRUE, return a list of replicate data frames. IfFALSE, return manifest tables only.
Value
An object of class mfrm_resamples with:
samples: replicate data frames with original Person, draw-unit, stratum and top-up metadata, or an empty list whenkeep_data = FALSE.manifest: replicate seeds, sampling scheme, base/added/final cluster counts, original Person counts, rating-row counts and coverage checks.stratum_manifest:BasePersonClustersbefore top-up,TopupPersonClusters, finalSelectedPersonClustersandSelectedOriginalPersons, and finalRowsfor each stratum.preserve_manifest: observed coverage of requested facet levels.specandterminology: inputs, sampling assumptions and interpretation.summary()retains full manifests as well as compact tables. For older results with top-up, pre-top-up stratum counts are retained as base counts; unrecorded final cluster counts areNA, with a note, while final row counts remain available. This does not redraw old samples.
Details
See build_mfrm_resampling_spec() for sampling assumptions,
top-up, label handling and saved-specification compatibility. Sampling
keeps each selected Person's entire rating cluster and re-keys every
instance when drawing with replacement. Other facet labels are retained.
Examples
# Same-size Person-cluster bootstrap within the observed Group strata.
ratings <- load_mfrmr_data("example_operational")
spec <- build_mfrm_resampling_spec(
ratings, person = "Person", facets = c("Rater", "Criterion"),
score = "Score", strata = "Group", reps = 2,
design = "stratified_bootstrap", sample_fraction = 1,
topup_preserve_facets = FALSE, seed = 99
)
draws <- draw_mfrm_resamples(spec, keep_data = FALSE)
summary(draws)$overview
#> Design SamplingScheme Reps SamplesReturned Replace
#> 1 stratified_bootstrap stratified_person_bootstrap 2 0 TRUE
#> PreserveCoverageCompleteRate TopupReps GapOrFallbackReps
#> 1 1 0 0
