Skip to contents

Identify Problematic States

identify_problematic_states.Rd

Identify states with estimation problems based on diagnostic information. Supports DCC, CGARCH, and GOGARCH models with model-specific checks.

Usage

identify_problematic_states(
  diagnostics,
  state = NULL,
  model_type = c("auto", "dcc", "cgarch", "gogarch")
)

Arguments

diagnostics

An object of class ms_diagnostics

state

Integer state index. If NULL (default), checks all states.

model_type

Character: model type override. One of "auto" (default), "dcc", "cgarch", or "gogarch". When "auto", attempts to detect from diagnostics.

Value

List with:

has_problems

Logical indicating if any problems were found

n_states_affected

Number of states with problems

problems

Named list with problem descriptions per state

Details

The function performs different checks depending on model type:

DCC

  • High persistence (alpha + beta > 0.98)

  • Constant correlation fallback

  • Parameter instability in final iterations

  • Boundary events

CGARCH

  • All DCC checks

  • Copula shape parameter issues (MVT: df < 3 or > 100)

  • ADCC gamma constraints

  • PIT transformation warnings

GOGARCH

  • ICA convergence failure

  • Mixing matrix ill-conditioning (condition number > 1000)

  • Unmixing matrix near-singularity

  • Component correlation (should be < 0.2)

  • Component GARCH high persistence

Examples

if (FALSE) { # \dontrun{
# Check all states with auto-detection
problems <- identify_problematic_states(diag)
if (problems$has_problems) {
  print(problems$problems)
}

# Check specific state for CGARCH model
problems <- identify_problematic_states(diag, state = 1, model_type = "cgarch")
} # }