Base Classes

Base Classes#

Abstract base classes for spey statistical-model backends.

This module defines two abstract base classes:

  • BackendBase — the interface every statistical-model backend must implement to integrate with spey’s hypothesis-testing machinery.

  • ConverterBase — a lightweight interface for objects that convert one statistical model representation into a BackendBase instance.

For a step-by-step guide on writing, registering, and packaging a new backend see the Building a plugin tutorial in the documentation.

backend_base.BackendBase()

Abstract base class that every spey statistical-model backend must inherit.

backend_base.ConverterBase()

Abstract base class for objects that convert one statistical model into another.

model_config.ModelConfig(poi_index, ...[, ...])

Container to hold certain properties of the backend and statistical model.

Hypothesis Testing Base#

Abstract base class for hypothesis testing in spey.

This module defines HypothesisTestingBase, which provides the complete hypothesis-testing API built on top of any statistical model backend. StatisticalModel inherits from this class, so every backend automatically receives the following capabilities as soon as it implements the two mandatory methods (config() and get_logpdf_func()):

  • \(\chi^2\) test statistics and profile likelihood ratio computation;

  • exclusion confidence levels (\(CL_s\)) via asymptotic, toy, or \(\chi^2\) calculators;

  • one-sided and two-sided \(\chi^2\) interval finding;

  • POI upper limits at arbitrary confidence levels;

  • discovery significance (\(\sqrt{q_0}\)).

HypothesisTestingBase([ntoys])

Abstract base class that provides the full hypothesis-testing API for spey.