smooth.ES.vcov

ES.vcov(type=None, bootstrap=False, heuristics=None, step_size=None, **boot_kwargs)

Variance-covariance matrix of the estimated parameters.

Mirrors R’s vcov.adam. type selects the estimator: "opg" (the default, the OPG/BHHH covariance built from the per-observation scores, PSD by construction and finite at boundary estimates), "hessian" (the observed Fisher Information) or "bootstrap".

Parameters:

  • type ({"opg", "hessian", "bootstrap"}, optional) – Covariance estimator; defaults to "opg".

  • bootstrap (bool, default=False) – Deprecated alias for type="bootstrap" (warns).

  • heuristics (float, optional) – If given, returns diag(abs(coef) * heuristics).

  • step_size (float, optional) – Finite-difference step.

  • **boot_kwargs – Forwarded to coefbootstrap() for the bootstrap type.

Returns:

Covariance matrix indexed/columned by coef_names.

Return type:

pandas.DataFrame

Parent Class: ES