Analyze the MFRM design network

Usage

mfrm_network_analysis(
  fit = NULL,
  diagnostics = NULL,
  top_n_subsets = NULL,
  min_observations = 0,
  include_graph = FALSE,
  data = NULL,
  person = NULL,
  facets = NULL,
  score = NULL,
  basis = c("observed", "planned"),
  weight = NULL
)

Arguments

fit: Output from fit_mfrm().
diagnostics: Optional output from diagnose_mfrm().
top_n_subsets: Optional maximum number of connected-subset rows to retain before constructing the graph.
min_observations: Minimum observations required to keep a subset row.
include_graph: Logical; if TRUE, include the underlying igraph object in the returned bundle. Defaults to FALSE so outputs remain easy to serialize.
data: Optional raw assignment/rating data, supplied instead of fit.
person, facets: Names of the person column and one or more facet columns in data. Include only Rater in facets to inspect person-rater links without a common Criterion node obscuring separate panels.
score: Numeric score column for basis = "observed". Recode study-specific missing markers to NA first; no category recoding or estimation is done.
basis: "observed" includes rows with complete identifiers and nonmissing finite scores; "planned" includes complete assignment rows without scores. Planned networks require score = NULL and weight = NULL.
weight: Optional numeric weight column for observed raw data. Missing and zero-weight rows do not create edges. Edge weights still count rows, not sampling weights. Negative or infinite weights are refused.

Value

A bundle of class mfrm_network_analysis containing:

summary: graph-level connectedness and vulnerability metrics
node_metrics: node-level degree, strength, graph-theoretic centrality, and cutpoint flags
edge_metrics: edge-level weights, betweenness, and bridge flags
facet_summary: facet-level aggregation of node/bridge indicators
cut_nodes: articulation-point rows from node_metrics
bridge_edges: bridge rows from edge_metrics

Details

mfrm_network_analysis() treats the person/facet-level observation design as an undirected weighted graph. Nodes are person or facet levels; edges connect levels that co-occur in at least one observed rating; edge weights are co-observation counts. The resulting network metrics are design diagnostics, not psychometric measures of person ability or rater quality. This is the package's assignment/co-observation network: unlike the score-relation networks returned by rater_network_analysis() and rater_halo_network_analysis(), it exposes disconnected co-observation components. Graph connectedness alone does not establish model identification. SourceSubsets, RetainedSubsets and ScopeComplete record whether subset filters retained the full observed design. Connectedness of selected subsets cannot be generalized to omitted subsets. Recreate older design reviews from the existing fit and matching diagnostics to record this coverage. plot(net, type = "centrality"), plot(net, type = "facet_summary"), and plot(net, type = "network") provide immediate visual checks; use draw = FALSE to extract reusable plot data.

The most useful review columns are:

Components: more than one component means separate co-observation regions, including known levels without observed links.
IsArticulationPoint: a node whose removal would increase disconnectedness.
IsBridge: an edge whose removal would increase disconnectedness.
Betweenness: a routing-dependence indicator; high values identify levels that carry many shortest paths through the design graph.

In incomplete rater-mediated designs, these graph summaries help identify fragile linking structures before interpreting facet measures or planning additional data collection.

References

McEwen, M. R. (2015). Development of a Software Prototype for Generating and Classifying Incomplete Many-Facet-Rasch Model Rating Designs. Brigham Young University.
Csardi, G., Nepusz, T., Traag, V., Horvat, S., Zanini, F., Noom, D., & Muller, K. (2026). igraph: Network Analysis and Visualization.

Before fitting

Supply data, person, facets and score instead of fit. Known levels from incomplete/unscored rows remain as nodes, including isolates. A planned assignment roster can be examined with basis = "planned"; absent assignments cannot be reconstructed from an observed-only file. row_usage records each input row's inclusion and first exclusion reason. With a fit, only its retained rows are available here; review original missing assignments before fitting. No prefit result supplies an identification audit or model-readiness decision. settings$is_disjoint describes the source graph, even when subset filters omit components; compare source/retained subset counts. Filters do not establish linking. See online networks and reporting guide for a panel example.

Examples

# Supplementary workflow: run explicitly to keep routine checks brief.
# example("mfrm_network_analysis", run.dontrun = TRUE)
if (FALSE) { # \dontrun{
ratings <- load_mfrmr_data("example_operational")
if (requireNamespace("igraph", quietly = TRUE)) {
  net <- mfrm_network_analysis(data = ratings, person = "Person",
    facets = "Rater", score = "Score")
  net$summary
  plot(net, type = "network", draw = FALSE)
}
} # }