quantlab.risk.base.ReturnCovariance

class quantlab.risk.base.ReturnCovariance(*args, **kwargs)

Bases: Protocol

The covariance of one-bar returns at a bar, over the symbols it covers: what a covariance estimator returns.

What a rule holding a covariance estimator may ask of its estimate, whatever the estimator. Three types satisfy it and stay separate types (ADR 0028): the dense quantlab.portfolio.base.CovarianceEstimate, the low-rank quantlab.portfolio.base.LowRankCovarianceEstimate (Ledoit-Wolf) and a factor risk model’s FactorRiskForecast. A rule asks the estimate and does not tell the types apart.

Attributes:

symbolsnp.ndarray

The n symbols covered, in the order of every per-symbol array.

covariancenp.ndarray

The dense [n, n] covariance.

variancenp.ndarray

Each symbol’s variance, [n].

Examples

>>> forecast = FactorRiskForecast(
...     symbols=np.array(["AAA", "BBB"]),
...     factor_names=("market",),
...     exposures=np.array([[1.0], [0.5]]),
...     factor_covariance=np.array([[0.04]]),
...     specific_variance=np.array([0.01, 0.02]),
... )
>>> isinstance(forecast, ReturnCovariance)
True

factor_exposure(name, symbols)

Return each of symbols’ exposure to risk factor name, [len(symbols)].

NaN for a symbol the estimate does not cover. An estimate without risk factors raises ValueError.

factor_form()

Return (B, F, D) when the covariance is B F B' + diag(D), else None.

From the factor form an optimiser prices risk without the dense matrix.

scaled(factor)

Return the estimate with every variance multiplied by factor (one bar to factor bars).

subset(rows)

Return the estimate over the symbols at positions rows, in that order.