skip to navigation
skip to content

ABBA 0.1.0

Tools for statistical analysis of A/B test results

Tools for statistical analysis of A/B test results.

ABBA provides several statistical tools for analysis of binomial data, typically resulting from A/B tests:

  • Wald and Agresti-Coull confidence intervals on binomial proportions
  • Confidence intervals on the difference and ratio of two binomial proportions
  • Hypothesis tests for inequality of two binomial proportions
  • Multiple test correction for control of familywise error rate

Some simple example usage:

>>> import abba.stats
>>> abba.stats.confidence_interval_on_proportion(
...     num_successes=50, num_trials=200, confidence_level=0.99)
ValueWithInterval(value=0.25, lower_bound=0.17962262748069852, upper_bound=0.33643200973247306)

>>> experiment = abba.stats.Experiment(
...     num_trials=5, baseline_num_successes=50, baseline_num_trials=200)
>>> results = experiment.get_results(num_successes=70, num_trials=190)
>>> results.relative_improvement
ValueWithInterval(value=0.4736842105263157, lower_bound=-0.014130868125315277, upper_bound=0.90421878236700903)
>>> results.two_tailed_p_value

ABBA requires SciPy for underlying statistical functions.

For more info, see the docstrings, unit tests, and the ABBA website (including an interactive Javascript version) at

File Type Py Version Uploaded on Size
ABBA-0.1.0.tar.gz (md5) Source 2012-09-28 6KB