Crate prospicio_prob 

Risk mathematics: the hub every domain reads and writes.

Holds the Distribution and Severity traits, the parametric Lognormal, the sampled representation Sampled, the shared risk measures, Distortion risk measures and their allocation, copulas, extreme value tails (evt) and PredictiveDistribution, the joint result every model returns. The discretized representation follows docs/design/distributions.md.

With the arrow feature, PredictiveDistribution reads and writes Arrow IPC files; the format is described in ipc.

Re-exports

pub use copula::Archimedean;

pub use copula::ArchimedeanCopula;

pub use copula::Copula;

pub use copula::GaussianCopula;

pub use copula::StudentTCopula;

pub use counting::Binomial;

pub use counting::Counting;

pub use counting::NegativeBinomial;

pub use counting::PanjerClass;

pub use counting::Poisson;

pub use custom::Custom;

pub use dist::Dist;

pub use dist::SeverityDist;

pub use distortion::Distortion;

pub use distribution::Distribution;

pub use gamma::Gamma;

pub use grid::Discretization;

pub use grid::DiscretizationReport;

pub use grid::Grid;

pub use large_losses::LargeLosses;

pub use local_pareto::LocalParetoApproximation;

pub use local_pareto::LocalParetoConversion;

pub use local_pareto::LogAffinePareto;

pub use local_pareto::local_pareto_to_piecewise;

pub use loglogistic::Loglogistic;

pub use lognormal::Lognormal;

pub use mixture::Mixture;

pub use pareto::Pareto;

pub use piecewise_pareto::PiecewisePareto;

pub use piecewise_pareto::Truncation;

pub use predictive::ComponentKey;

pub use predictive::KeyValue;

pub use predictive::PredictiveDistribution;

pub use provenance::InputHasher;

pub use provenance::Provenance;

pub use sampled::Empirical;

pub use sampled::Sampled;

pub use severity::Severity;

pub use tweedie::Tweedie;

pub use weibull::Weibull;

Modules

capital

Capital allocation and diversification for a joint PredictiveDistribution.

copula

Copulas: dependence between marginals, separate from the marginals.

counting

Claim-count (frequency) distributions on 0, 1, 2, ….

custom

Custom, a user-defined loss severity given by its cdf: the “slow path” of docs/design/distributions.md, through which a Python or R function enters the native calculations.

dist

Dist, the closed enum of every native distribution, for the places where the family is chosen at run time: the Python and R bindings, serialization and model outputs (docs/design/distributions.md, “Static vs dynamic dispatch”).

distortion

Distortion risk measures: ρ(X) = ∫ g(S(x)) dx for a concave distortion g of the survival function, with g(0) = 0, g(1) = 1.

distribution

The trait shared by every distribution representation.

evt

Extreme value tails: a generalized Pareto distribution (GPD) fitted to the exceedances over a threshold, for VaR and TVaR beyond the draws.

gamma

The gamma distribution.

grid

The discretized representation: probabilities on an evenly spaced grid.

large_losses

Large-loss data with the two usual defects, and maximum likelihood fits of Pareto-family severities to it (see docs/design/pareto.md).

local_pareto

The log-affine local Pareto distribution: a local Pareto alpha that rises linearly in the log of the amount (see docs/design/pareto.md).

loglogistic

The loglogistic (Fisk) distribution.

lognormal

The lognormal distribution.

mixture

Finite mixtures of severities.

pareto

The single-parameter (European) Pareto distribution, optionally truncated: the standard large-loss severity in reinsurance pricing (see docs/design/pareto.md).

piecewise_pareto

Piecewise Pareto: a different Pareto alpha above each threshold, the output of tower matching and the general large-loss model in docs/design/pareto.md.

portfolio

Building a portfolio from predictive distributions of different models: a reserve bootstrap by origin, premium risk by line, a reinsurance tower’s gross, ceded and net.

predictive

The joint predictive distribution every model returns.

provenance

Where a result came from: model, parameters, seed, versions and a hash of the input.

risk

Risk measures on sorted draws, and the exponential-utility measures (entropic, Esscher) on any draws.

sampled

The sampled representation: a distribution known only through draws.

serial

Saving and loading distributions: Dist::to_json and Dist::from_json, a versioned JSON document per distribution.

severity

Operations on a loss severity that are exact for parametric (and later discretized) distributions: limited expected values, stop-loss and layers.

tweedie

The Tweedie distribution with power 1 < p < 2: compound Poisson with gamma severities.

weibull

The Weibull distribution.