Skip to contents

Alternative to estimate_garch_weighted_gogarch using improved_ica_decomposition for better convergence.

Usage

estimate_gogarch_improved(
  residuals,
  weights,
  spec,
  ica_restarts = 3,
  diagnostics = NULL,
  verbose = FALSE
)

Arguments

residuals

T x k matrix of residuals

weights

T-vector of observation weights

spec

Model specification (see estimate_garch_weighted_gogarch)

ica_restarts

Number of ICA restarts (default: 3)

diagnostics

Optional diagnostics collector

verbose

Print progress

Value

Same structure as estimate_garch_weighted_gogarch with ICA quality metrics in coefficients$ica_info$quality

Details

Use this when ICA convergence is problematic. Provides warnings when ICA quality is poor (independence score < 0.8) and falls back to PCA if ICA fails completely.