Compute Standard Errors for ADCC Parameters — adcc_standard

Computes Hessian-based (or sandwich) standard errors for the 3-parameter ADCC model: (alpha, gamma, beta), optionally with shape for MVT.

Usage

adcc_standard_errors(
  params,
  z_matrix,
  weights,
  Qbar,
  Nbar = NULL,
  copula_dist = "mvn",
  method = c("hessian", "sandwich")
)

Arguments

params: Parameter vector: c(alpha, gamma, beta) or c(alpha, gamma, beta, shape)
z_matrix: T x k matrix of copula residuals (PIT-transformed)
weights: T-vector of observation weights
Qbar: k x k unconditional correlation matrix
Nbar: k x k average outer product of negative residuals (optional)
copula_dist: "mvn" or "mvt"
method: SE method: "hessian" (default) or "sandwich"

Value

List with:

se: Standard errors for each parameter
vcov: Variance-covariance matrix
hessian: Hessian matrix of NLL
eigenvalues: Eigenvalues of Hessian (for diagnostics)
condition_number: Condition number of Hessian
param_names: Parameter names
method: "hessian" or "sandwich"
valid: Logical: are SEs valid?

Details

For ADCC, the gamma parameter introduces asymmetric correlation dynamics. The Hessian-based SEs may underestimate uncertainty for high-persistence models (similar to standard DCC). Bootstrap SEs are recommended for publication-quality inference.