Current version: 2021-07-24
- Experimenter can now evaluate a basic experiment results dataset with BinomialExperiment.ingest_data()
- Still designed for experiments with binary outcomes (follow a binomial distribution)
In completing this project, my primary motivation is learning the ins and outs of statistical hypothesis testing and exploring the relationships among statistical power, effect size and sample size.
The Jupyter notebooks and models were originally created according to that motivation.
A second motivation has since emerged: creating a web app that allows a marketer to:
- Test the results of split tests for statistical significance
- Plan such split tests by determining how large a test sample needs to be in order to detect a desired effect size.
- Explore confidence intervals when they are more insightful or easier to communicate than significance
This is a personal project that I started while enrolled as a student in Udacity's Data Scientist Nanodegree program.
I've written all code and designed all programs, here. However, I've been influenced by methods and techniques demonstrated by my instructors at Udacity.