API Reference¶

Full API documentation for all bayesprop modules, auto-generated from source code docstrings.

Modules¶

Module	Description
Data Schemas	Pydantic data contracts — `HypothesisDecision`, `ROPEResult`, `SavageDickeyResult`, configs
Non-Paired Model	Independent Beta-Bernoulli A/B test
Paired Model (Laplace)	Paired logistic model with Laplace approximation
Paired Model (Pólya-Gamma)	Paired logistic model with PG Gibbs sampler
Paired Model (Bayesian Bootstrap)	Nonparametric paired test via Dirichlet weights on the empirical distribution
Sequential designs	Warm-started sequential variants of the non-paired and paired-Laplace models
Utilities	DGPs, frequentist baselines, BFDA, decision helpers
Operating Characteristics	Monte-Carlo OC simulation harness for the non-paired model (fixed-n + sequential)
Operating Characteristics (Paired)	Monte-Carlo OC simulation harness for the paired-Laplace model (fixed-n + sequential)

Shared decision-rule interface¶

All four model classes inherit from BaseBayesPropTest and expose the same core workflow:

Fit → Summarise → Decide¶

Method	Return type	Description
`model.fit(y_A, y_B)`	`self`	Fit the model to observed data (method chaining)
`model.summary`	`PairedSummary` / `NonPairedSummary`	Posterior summary (θ means, Δ, CI, P(A > B))
`model.decide(rule=None)`	`HypothesisDecision`	Run BF + P(H₀) + ROPE in a single call
`model.rope_test(rope=None, ci_mass=0.95)`	`ROPEResult`	ROPE analysis on the difference posterior
`model.print_summary()`	`None`	Print a human-readable summary to stdout

Plotting¶

Method	Description
`model.plot_posteriors()`	Single-panel overlay of θ_A and θ_B posteriors
`model.plot_posterior_delta()`	Single-panel KDE of Δ = θ_A − θ_B (probability scale) with 95 % CI

Parametric-only methods¶

The following are available on the three parametric models (NonPairedBayesPropTest, PairedBayesPropTest(method="laplace"), PairedBayesPropTest(method="pg")) but not on the Bayesian bootstrap (PairedBayesPropTest(method="bootstrap")), which has no parametric prior on Δ:

Method	Return type	Description
`model.savage_dickey_test(null_value=0.0)`	`SavageDickeyResult`	Savage-Dickey Bayes factor at the point null
`model.posterior_probability_H0(BF_01, prior_H0=0.5)`	`PosteriorProbH0Result`	Static: convert a BF₀₁ to posterior P(H₀ ∣ D)
`model.plot_savage_dickey()`	`None`	Prior vs posterior density with BF annotation