Mean Models

All ARCH models start by specifying a mean model.

ZeroMean([y, hold_back, volatility, ...])

Model with zero conditional mean estimation and simulation

ConstantMean([y, hold_back, volatility, ...])

Constant mean model estimation and simulation.

ARX([y, x, lags, constant, hold_back, ...])

Autoregressive model with optional exogenous regressors estimation and simulation

HARX([y, x, lags, constant, use_rotated, ...])

Heterogeneous Autoregression (HAR), with optional exogenous regressors, model estimation and simulation

LS([y, x, constant, hold_back, volatility, ...])

Least squares model estimation and simulation

(G)ARCH-in-mean Models

(G)ARCH-in-mean models allow the conditional variance (or a transformation of it) to enter the conditional mean.

ARCHInMean([y, x, lags, constant, ...])

(G)ARCH-in-mean model and simulation

Special Requirements

Not all volatility processes support application to AIM modeling. Specifically, the property updateable must be True.

In [1]: from arch.univariate import GARCH, EGARCH

In [2]: GARCH().updateable
Out[2]: True

In [3]: EGARCH().updateable
Out[3]: True

Writing New Mean Models

All mean models must inherit from :class:ARCHModel and provide all public methods. There are two optional private methods that should be provided if applicable.

ARCHModel([y, volatility, distribution, ...])

Abstract base class for mean models in ARCH processes.