Discover candidate fiscal-health dimensions — fiscal

Uses PCA or exploratory factor analysis to propose empirical dimensions from indicators retained by fiscal_triage(). This function discovers and describes candidate groupings; it does not name dimensions, orient scores, or construct a composite fiscal-health index.

Usage

fiscal_dimensions(
  data,
  triage = NULL,
  variables = NULL,
  method = c("pca", "efa"),
  nfactors = NULL,
  rotate = "oblimin",
  fm = "minres",
  missing = c("complete", "median"),
  primary_loading = 0.4,
  cross_loading = 0.3,
  min_indicators = 2L,
  parallel_iter = 100L,
  seed = NULL
)

Arguments

data: A data frame containing the candidate indicators.
triage: An optional fiscal_triage object. When supplied, only its selected_for_dimensions indicators enter the model and its standalone indicators are carried into the result.
variables: Optional character vector of indicators. Supply either triage or variables, not both.
method: Dimension-discovery method: principal components analysis ("pca") or exploratory factor analysis ("efa").
nfactors: Number of dimensions. If NULL, parallel analysis is used.
rotate: Rotation passed to psych::principal() or psych::fa().
fm: Factoring method used when method = "efa".
missing: Missing-data rule. "complete" uses complete rows; "median" replaces missing values with variable medians for this model.
primary_loading: Minimum absolute loading for assignment to a candidate dimension.
cross_loading: Minimum absolute secondary loading used to flag a cross-loading indicator.
min_indicators: Minimum assigned indicators for a proposed cluster to be treated as a multi-indicator dimension rather than a singleton.
parallel_iter: Number of simulated datasets used by parallel analysis when nfactors is NULL.
seed: Optional seed for reproducible parallel analysis.

Value

An object of class fiscal_dimensions containing the fitted model, loading and assignment tables, proposed clusters, singleton and unassigned indicators, triage standalones, and the complete modeling specification.