mmc_lev: MMC Test for Leverage in SV(p) Models

In wARMASVp: Winsorized ARMA Estimation for Higher-Order Stochastic Volatility Models

MMC Test for Leverage in SV(p) Models

Description

Performs a Maximized Monte Carlo (MMC) test of the null hypothesis H_0: \rho = \rho_0 by maximizing the MC p-value over nuisance parameters (phi, sigma_y, sigma_v).

Usage

mmc_lev(
  y,
  p = 1,
  J = 10,
  N = 99,
  rho_null = 0,
  burnin = 500,
  eps = NULL,
  threshold = 1,
  method = "pso",
  maxit = NULL,
  rho_type = "pearson",
  del = 1e-10,
  trunc_lev = TRUE,
  wDecay = FALSE,
  Bartlett = FALSE,
  Amat = NULL,
  errorType = "Gaussian",
  logNu = FALSE,
  sigvMethod = "factored",
  winsorize_eps = 0
)

Arguments

y	Numeric vector. Observed returns.
p	Integer. Order of the volatility process. Default 1.
J	Integer. Winsorizing parameter. Default 10.
N	Integer. Number of Monte Carlo replications. Default 99.
rho_null	Numeric. Value of \rho under the null. Default 0.
burnin	Integer. Burn-in for simulation. Default 500.
eps	Numeric vector. Half-width of the search region around the estimated nuisance parameters. For Gaussian: length p+2 (phi, sigma_y, sigma_v). For Student-t/GED: length p+2 (phi, sigma_y, sigma_v; nu bounds set proportionally at +/-30 length p+3 (phi, sigma_y, sigma_v, nu). Default NULL which uses rep(0.3, p+2) with proportional nu bounds.
threshold	Numeric. Target p-value (optimization stops if reached). Default 1.
method	Character. Optimization method: "pso" (particle swarm), "GenSA" (generalized simulated annealing). Default "pso".
maxit	Integer or list. Maximum iterations/evaluations for the optimizer. Default depends on method.
rho_type	Character. Correlation type. Default "pearson".
del	Numeric. Small constant for log transformation. Default 1e-10.
trunc_lev	Logical. Truncate leverage correlation estimate to [-0.999, 0.999]. Default TRUE.
wDecay	Logical. Use decaying weights. Default FALSE.
Bartlett	Logical. If TRUE, use Bartlett kernel HAC weighting matrix. If FALSE, use identity matrix. Default FALSE.
Amat	Weighting matrix specification. NULL (default) for identity weighting, "Weighted" for data-driven HAC, or a numeric matrix of dimension (p+3)x(p+3) (Gaussian) or (p+4)x(p+4) (heavy-tail). Takes precedence over Bartlett.
errorType	Character. Error distribution: "Gaussian" (default), "Student-t", or "GED".
logNu	Logical. Use log-space for nu estimation (Student-t only). Default FALSE.
sigvMethod	Method for sigma_v estimation: "factored" (default), "direct", or "hybrid".
winsorize_eps	Number of extreme autocovariance lags to winsorize (0 = none). Default 0.

Value

A list with the optimization output including:

value

Maximized p-value.

par

Nuisance parameter values at the maximum.

Additional fields depend on the optimization method used.

Examples

y <- sim_svp(1000, phi = 0.95, sigy = 1, sigv = 0.2, leverage = TRUE, rho = -0.3)$y
mmc <- mmc_lev(y, p = 1, J = 10, N = 19, method = "pso", maxit = 10)
mmc$value

wARMASVp documentation built on May 15, 2026, 5:07 p.m.