pymgarch

Multivariate GARCH for Python: DCC, ADCC, and CCC correlation dynamics on top of arch univariate marginals, validated against R’s rmgarch.

Python has had no maintained general-purpose multivariate GARCH framework: the existing packages stop at Gaussian DCC(1,1), while R users have had a full model stack in rmgarch (now tsmarch) for a decade. pymgarch closes that gap incrementally. Stage 1 (univariate volatility) is delegated to arch; stage 2 (correlation dynamics) is what this library implements, with Engle-Sheppard two-stage standard errors and replication tests against rmgarch’s fitted parameters and likelihoods.

• Quickstart
  • Distributions and asymmetry
  • Custom marginals
  • Forecasting and filtering
  • Scaling note
• Examples
  • Fit, inspect, forecast
  • Model comparison
  • Custom marginals and filtering
• Models
  • CCC (Bollerslev 1990)
  • DCC(1,1) (Engle 2002)
  • ADCC (Cappiello, Engle and Sheppard 2006)
  • GO-GARCH (van der Weide 2002)
  • Copula-GARCH (Patton 2006)
  • Composite likelihood for large N
  • Scalar and diagonal BEKK (Engle and Kroner 1995)
  • Forecasting
  • Reported likelihood
• Inference
  • Two-stage standard errors
  • Known approximations
  • Replication against rmgarch
• API reference
  • Models
  • Marginals
  • Results
  • Simulation

Install

pip install pymgarch          # or: pip install pymgarch[numba]