QUANTA - LAB GARCH

Institutional volatility modeling suite with GARCH estimation, VaR/CVaR risk metrics, and Basel III backtesting.
Models Available:

GARCH(1,1) — symmetric volatility clustering
GJR-GARCH(1,1) — asymmetric leverage effect
EGARCH(1,1) — log-variance specification

Risk Metrics:

VaR (95%/99%) with Student-t fat tails
CVaR/Expected Shortfall (coherent risk measure)
Multi-horizon VaR (1d, 5d, 10d) with persistence-adjusted scaling
DoF estimation via method of moments (±15-25% uncertainty)

Backtesting (Basel III Compliant):

Kupiec unconditional coverage test
Christoffersen independence test
Traffic light system (Green/Yellow/Red zones)

Diagnostics:

ARCH-LM test for residual effects
AIC/BIC information criteria
Structural break detection (CUSUM-based)
Jump/outlier detection
Model confidence score (0-100)

V3.6 Improvements:

Adaptive grid search (~60% faster)
High persistence warning (p > 0.98)
Persistence-adjusted multi-horizon scaling (better than √T)

Dashboard Includes:

Real-time conditional volatility (annualized)
Parameter estimates (α, β, γ, θ)
Persistence and half-life
Regime classification (Normal/Elevated/Crisis)

Important:

Grid search produces point estimates (no confidence intervals)
Parameters may differ ±3-5% from true MLE
NOT for illiquid assets or significant overnight gaps
Screening tool only — validate with Python arch / R rugarch

References: Bollerslev (1986), Nelson (1991), GJR (1993), Engle (1982), McNeil et al. (2015), Kupiec (1995), Christoffersen (1998)