SV.fit#

SV.fit(*, method='qml', n_particles=500, seed=None)[source]#

Point-estimate the model.

Parameters:

  • method (str) – "qml" for Harvey-Ruiz-Shephard quasi-maximum likelihood on the linearized model, "particle" for simulated maximum likelihood on the particle filter’s estimate under common random numbers (Nelder-Mead from the QML start).

  • n_particles (int) – Particles per evaluation under "particle", at least 50.

  • seed (int | None) – The common seed under "particle".

Return type:

SVResult

Returns:

The fitted SVResult.

Raises:

SpecificationError – If the method is unknown, or n_particles is below 50 under "particle".

Example

QML is the fast estimator; the particle estimator starts from it and is far slower, so it is not run here:

>>> import numpy as np
>>> from cultivars.engine._internals._simulators import _simulate_stochastic_volatility
>>> rng = np.random.default_rng(0)
>>> y, _ = _simulate_stochastic_volatility(
...     300, mu=-1.0, phi=0.95, sigma2=0.05, mean=0.0, rng=rng
... )
>>> res = SV(y, mean="zero").fit()
>>> res.method, bool(0.85 < res.phi < 1.0)
('qml', True)
>>> SV(y).fit(method="mle")
Traceback (most recent call last):
    ...
cultivars.exceptions.SpecificationError: method must be 'qml' or 'particle'; got 'mle'.
On this page