Audit empirical identifiability from replicate estimates — audit

Flags parameters with near-zero estimate variance, explosive dispersion, excessive missingness, or strongly correlated estimates when a covariance matrix is supplied.

Usage

audit_identifiability(
  results,
  max_missing = 0.05,
  max_sd_ratio = 10,
  correlation_matrix = NULL,
  max_abs_correlation = 0.995
)

Arguments

results: Validation or model results.
max_missing: Maximum acceptable missingness.
max_sd_ratio: Maximum acceptable standard-deviation ratio.
correlation_matrix: Optional parameter-correlation matrix.
max_abs_correlation: Maximum acceptable absolute parameter correlation.

Value

An object of class "eye_irt_identifiability_audit", "data.frame", stored as a data frame, containing empirical identifiability from replicate estimates and associated metadata needed to interpret the result.