GARCHResult.unconditional_variance

property GARCHResult.unconditional_variance: float | None

The long-run variance omega / (1 - persistence).

Returns:

None for the log-variance family, whose stationary variance is the mean of a lognormal and has no closed form in the reported parameters, and None when the process is not covariance stationary. Returning a number in either case would invite it straight into a risk calculation.

Example

>>> rng = np.random.default_rng(0)
>>> y = rng.standard_normal(500)
>>> res = GARCH(y, p=1, q=1).fit()
>>> bool(np.isclose(res.unconditional_variance, res.omega / (1 - res.persistence)))
True