expan 0.5.1

Experiment Analysis Library

A/B tests (a.k.a. Randomized Controlled Trials or Experiments) have been widely applied in different industries to optimize business processes and user experience. ExpAn (Experiment Analysis) is a Python library developed for the statistical analysis of such experiments and to standardise the data structures used.

The data structures and functionality of ExpAn are generic such that they can be used by both data scientists optimizing a user interface and biologists running wet-lab experiments. The library is also standalone and can be imported and used from within other projects and from the command line.

Major statistical functionalities include:

• feature check

• delta

• subgroup analysis

• trend

Installation

To install ExpAn, run this command in your terminal:

$ pip install expan

Usage

To use ExpAn in a project:

import expan

Some mock-up data:

from expan.core.experiment import Experiment
from expan.core.util import generate_random_data

exp = Experiment('B', *generate_random_data())
exp.delta()

Documentation

The latest stable version is 0.5.1.

ExpAn main documentation

ExpAn Description - details about the concept of the library and data structures.

ExpAn Introduction - a full jupyter (iPython) notebook. You can view it as slides with jupyter:

sh serve_intro_slides

License

The MIT License (MIT)

Copyright © [2016] Zalando SE, https://tech.zalando.com

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the “Software”), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED “AS IS”, WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

Change Log

Unreleased

Full Changelog

Implemented enhancements:

• Bad code duplication in experiment.py #81

• pip == 8.1.0 requirement #76

Fixed bugs:

• Experiment.sga() assumes features and KPIs are merged in self.metrics #87

• pctile can be undefined in Results.to\_json\(\) #78

Closed issues:

• Results.to_json() => TypeError: Object of type ‘UserWarning’ is not JSON serializable #77

• Rethink Results structure #66

Merged pull requests:

• new dataframe tree traverser in to_json() #92 (gbordyugov)

• updated requirements.txt to have ‘greater than’ dependencies instead … #89 (mkolarek)

• pip version requirement #88 (gbordyugov)

• Test #86 (s4826)

• merging in categorical binning #84 (gbordyugov)

• Add documentation of the weighting logic #83 (jbao)

• Early stopping #80 (jbao)

• a couple of minor cleanups #79 (gbordyugov)

• Merge to_json() changes #75 (mkolarek)

• Feature/early stopping #73 (jbao)

v0.4.5 (2017-02-10)

Full Changelog

Fixed bugs:

• Numbers cannot appear in variable names for derived metrics #58

Merged pull requests:

• Feature/results and to json refactor #74 (mkolarek)

• Merge to_json() and prob_uplift_over_zero changes #72 (mkolarek)

• regex fix, see https://github.com/zalando/expan/issues/58 #70 (gbordyugov)

v0.4.4 (2017-02-09)

Full Changelog

Implemented enhancements:

• Add argument assume_normal and treatment_cost to calculate_prob_uplift_over_zero() and prob_uplift_over_zero_single_metric() #26

• host intro slides (from the ipython notebook) somewhere for public viewing #10

Closed issues:

• migrate issues from github enterprise #20

Merged pull requests:

• Feature/results and to json refactor #71 (mkolarek)

• new to_json() functionality and improved vim support #67 (mkolarek)

v0.4.3 (2017-02-07)

Full Changelog

Closed issues:

• coverage % is misleading #23

Merged pull requests:

• Vim modelines #63 (gbordyugov)

• Feature/octo 1253 expan results in json #62 (mkolarek)

• 0.4.2 release #60 (mkolarek)

v0.4.2 (2016-12-08)

Full Changelog

Fixed bugs:

• frequency table in the chi square test doesn’t respect the order of categories #56

Merged pull requests:

• OCTO-1143 Review outlier filtering #59 (domheger)

• Workaround to fix #56 #57 (jbao)

v0.4.1 (2016-10-18)

Full Changelog

Merged pull requests:

• small doc cleanup #55 (jbao)

• Add comments to cli.py #54 (igusher)

• Feature/octo 545 add consolidate documentation #53 (mkolarek)

• added os.path.join instead of manual string concatenations with ‘/’ #52 (mkolarek)

• Feature/octo 958 outlier filtering #50 (mkolarek)

• Sort KPIs in reverse order before matching them in the formula #49 (jbao)

v0.4.0 (2016-08-19)

Full Changelog

Closed issues:

• Support ‘overall ratio’ metrics (e.g. conversion rate/return rate) as opposed to per-entity ratios #44

Merged pull requests:

• merging dev to master #48 (jbao)

• OCTO-825 overall metric #47 (jbao)

• Bump version: 0.3.2 → 0.3.4 #46 (mkolarek)

• Bug/fix dependencies #45 (mkolarek)

v0.3.4 (2016-08-08)

Full Changelog

Closed issues:

• perform trend analysis cumulatively #31

• Python3 #21

Merged pull requests:

• Feature/2to3 #43 (mkolarek)

v0.3.3 (2016-08-02)

Full Changelog

Merged pull requests:

• Merge pull request #41 from zalando/master #42 (jbao)

• master to dev #41 (mkolarek)

• Bump version: 0.3.1 → 0.3.2 #40 (mkolarek)

• Revert “Merge pull request #35 from zalando/dev” #39 (mkolarek)

• Merge pull request #35 from zalando/dev #38 (mkolarek)

v0.3.2 (2016-08-02)

Full Changelog

Merged pull requests:

• Bugfix/trend analysis bin label #37 (jbao)

• Added cumulative trends analysis OCTO-814 #36 (domheger)

• Merging 0.3.1 to master #35 (domheger)

v0.3.1 (2016-07-15)

Full Changelog

Merged pull requests:

• Bugfix/prob uplift over 0 #34 (jbao)

• Master #30 (mkolarek)

v0.3.0 (2016-06-23)

Full Changelog

Implemented enhancements:

• Add P(uplift>0) as a statistic #2

• Added function to calculate P(uplift>0) #24 (jbao)

Merged pull requests:

• updated travis.yml #29 (mkolarek)

• Master #28 (mkolarek)

• Master #27 (mkolarek)

• only store the p-value in the chi-square test result object #22 (jbao)

v0.2.5 (2016-05-30)

Full Changelog

Implemented enhancements:

• Implement __version__ #14

Closed issues:

• upload full documentation! #1

Merged pull requests:

• implement expan.__version__ #19 (pangeran-bottor)

• Mainly documentation changes, as well as travis config updates #17 (robertmuil)

• Update README.rst #16 (pangeran-bottor)

• added cli module #11 (mkolarek)

• new travis config specifying that only master and dev should be built #4 (mkolarek)

v0.2.4 (2016-05-16)

Full Changelog

Closed issues:

• No module named experiment and test_data #13

Merged pull requests:

• new travis config specifying that only master and dev should be built #5 (mkolarek)

v0.2.3 (2016-05-06)

Full Changelog

v0.2.2 (2016-05-06)

Full Changelog

v0.2.1 (2016-05-06)

Full Changelog

v0.2.0 (2016-05-06)

Merged pull requests:

• Added detailed documentation with data formats #3 (robertmuil)

* This Change Log was automatically generated by `github_changelog_generator <https://github.com/skywinder/Github-Changelog-Generator>`__