Ambrosia
Ambrosia is a Python library for A/B tests design, split and effect measurement. It provides rich set of methods for conducting full A/B test pipeline.
An experiment design stage is performed using metrics historical data which could be processed in both forms of pandas and spark dataframes with either theoretical or empirical approach.
Group split methods support different strategies and multi-group split, which allows to quickly create control and test groups of interest.
Final effect measurement stage is conducted via testing tools that are able to return relative and absolute effects and construct corresponding confidence intervalsfor continious and binary variables. Testing tools as well as design ones support significant number of statistical criteria, like t-test, non-parametric, and bootstrap.
For additional A/B tests support library provides features and tools for data preproccesing and experiment acceleration.
Key functionality
Pilots design ✈
Multi-group split 🎳
Matching of new control group to the existing pilot 🎏
Getting the experiments result evaluation as p-value, point estimate of effect and confidence interval 🎞
Experiments acceleration 🎢
Quickstart
Contents
Links