quantdata.dysanos.tangent_levy.metrics#
Static-fit, forecast, and no-arbitrage metrics for Tangent Lévy.
- class quantdata.dysanos.tangent_levy.metrics.PriceErrorMetrics(observations, weighted_rmse, rmse, mae, median_absolute_error, p95_absolute_error, p99_absolute_error, maximum_absolute_error, mean_signed_error)[source]#
Bases:
objectOutright errors against a smoothed normalized call target.
- class quantdata.dysanos.tangent_levy.metrics.IntervalErrorMetrics(observations, fraction_inside, fraction_outside, mean_all, median_all, mean_positive, median_positive, p95, p99)[source]#
Bases:
objectHalf-spread-normalized executable-interval diagnostics.
- class quantdata.dysanos.tangent_levy.metrics.NoArbitrageMetrics(lower_bound, upper_bound, strike_monotonicity, strike_convexity, calendar_monotonicity)[source]#
Bases:
objectMaximum finite-grid static-arbitrage violations.
- class quantdata.dysanos.tangent_levy.metrics.StatisticalFidelityMetrics(reference_observations, model_observations, dimensions, mean_change_error_norm, relative_covariance_error, correlation_error, median_variance_ratio, eigenvalue_spectrum_error, maximum_principal_angle_degrees, reference_covariance_condition, model_covariance_condition)[source]#
Bases:
objectCentered-innovation moment and factor-structure comparison.
- quantdata.dysanos.tangent_levy.metrics.price_error_metrics(model, target, *, half_spread=None, mask=None)[source]#
Compute original-strike or dense-grid outright price errors.
- quantdata.dysanos.tangent_levy.metrics.interval_error_metrics(model, bids, asks, *, half_spread_floor=0.0001, mask=None)[source]#
Implement the repository half-spread interval-error convention.
- quantdata.dysanos.tangent_levy.metrics.implied_volatility_error(*, tau, strikes, model_calls, target_calls, mask=None)[source]#
Return implied-volatility error summaries in volatility points.
- quantdata.dysanos.tangent_levy.metrics.no_arbitrage_metrics(calls, *, tau, strikes)[source]#
Measure normalized call bounds and finite-grid shape restrictions.
- quantdata.dysanos.tangent_levy.metrics.statistical_fidelity_metrics(reference_changes, model_changes, *, num_factors=5)[source]#
Compare two change samples without assigning a physical-measure claim.
- quantdata.dysanos.tangent_levy.metrics.energy_score(ensemble, observation, *, component_weights=None)[source]#
Compute the unbiased finite-ensemble multivariate energy score.