Workshop: An Introduction to Bayesian Statistics and Buy-Till-You-Die Models

For 2022, we are starting a series of workshops to introduce the concept of Bayesian statistics using a practical real-world problem: building customer analytics models to help predict customer behaviour.

All code for my workshops and talks can be found in this GitHub repo:

https://github.com/kaybenleroll/data_workshops

This workshop is contained in the ws_clvbayes_202201/ folder

This series of workshops is done live, with code being commited and pushed to the above GitHub repository as well as the Dockerfile infrastructure to build the container to run the code.

Bayesian statistics is an approach that enables the combination of domain knowledge, data and statistical modelling in a principled and logical way. The primary outputs of these models are distributions of the model parameters that allows us to quantify the uncertainty in our models.

In the third workshop of this series we fully focus on the use of Stan to fit a number of models of the transaction frequency. We discuss the use of priors, and why they are important, as well as some discussion around methods for validating the models. We will also look deeper into the various HMC-related diagnostics provided by Stan to help ensure the validity of the fit itself.

We strive to make our workshops as standalone as possible, so new attendees are always welcome!

This workshop will be broadcast using Zoom and will be recorded and made available online via our YouTube channel.