CPrior 0.4.0

CPrior 0.4.0 Release Notes

• Bayesian experiment class for A/B and multivariate testing.
• Functionalities to analyze and explore a Bayesian experiment throughout the process.

Utilities

• Exact computation of confidence/credible intervals using methods ETI and HDI. HDI implementation solves a constrained nonlinear programming problem using a SLSQP solver.

CPrior 0.3.1

CPrior 0.3.1 Release Notes

• Installation for Windows. Travis CI supports Windows build for Python 3.5, 3.6 and 3.7.
• CPrior available in PyPI.

Utilities

• Computation of confidence/credible intervals using methods ETI and HDI. All methods computing credible intervals for ABTest and MVTest are updated.

CPrior 0.3.0

CPrior 0.3.0 Release Notes

• Normal-inverse-gamma conjugate prior distribution.
• New Bayesian models supported: Normal-normal-inverse-gamma and Log-normal-normal-inverse-gamma model with unknown mean and variance.

Numerical methods

• Multivariate testing functions probability_vs_all, expected_loss_vs_all and expected_loss_relative_vs_all use numerical integration (method="quad") as default computational method. This provides a significant speed-up compared to other methods and more accurate solutions.
• Combination of asymptotic estimates and numerical integration for the marginal distribution of the mean in the normal-inverse-gamma conjugate prior distribution.

CPrior 0.2.0

CPrior 0.2.0 Release Notes

• Pareto conjugate prior distribution.
• New Bayesian model supported: Uniform-Pareto model.
• New multivariate testing method: expected_loss_relative_vs_all using MC or MLHS computation method.

CPrior 0.1.0

CPrior 0.1.0 Release Notes

• Bayesian A/B and Multivariate testing functionalities.
• Beta and gamma conjugate prior distributions.
• Bayesian models supported:
  • Beta: Bernoulli, binomial, geometric and negative binomial.
  • Gamma: exponential and Poisson.