risk.simulate()

Simulates marginals joined by a copula.

Usage

risk.simulate(
    copula,
    marginals,
    n_sims,
    seed,
    keys=None,
    dims=None,
)

In simulation i, draws uniforms from copula with stream i of seed and applies each marginal’s quantile function.

Parameters

copula: (GaussianCopula, StudentTCopula or ArchimedeanCopula)

marginals: list of Lognormal, Grid or Pareto-family severities

One per copula dimension.

n_sims: int

seed: int

keys: list of tuple = None

One component key per marginal; defaults to (0,), (1,), ....

dims: list of str = ["component"]

Returns

PredictiveDistribution

Examples

>>> from prospicio.distributions import Lognormal
>>> from prospicio.risk import GaussianCopula, simulate
>>> c = GaussianCopula([[1.0, 0.4], [0.4, 1.0]])
>>> pd = simulate(c, [Lognormal.from_mean_cv(100.0, 0.2), Lognormal.from_mean_cv(50.0, 1.0)],
...               10_000, 42, keys=[("motor",), ("property",)], dims=["lob"])
>>> abs(pd.mean() - 150.0) < 3.0

True