Bayesian methods are very useful when you have specific prior information and/or when you need to quantify
the uncertainty in your estimates. In this lab we’ll get into some situations where probabilities are useful outputs
from a model.
The dataset come from the following sources:
§ https://www.kaggle.com/zhangluyuan/ab-testing
Tasks:
- Suppose the marketing people are testing a new web page, with the hope of increasing the conversion rate
(proportion of visitors who sign up or take some other action). We'll imagine that you collected the data in the
ab_data.csv file which lists user visits with whether it was the new page or old page, and whether there was a
conversion. Explore and visualize the data. - Instead of using p-values, do a Bayesian AB test. You can define and update independent priors on the old and
new conversion rates and arrive at respective posterior distributions. Try a prior of Beta(alpha=2, beta=20) for
the old rate, which represents what has been observed in the past.