Bayesian Data Analysis: an Introductory Workshop

Workshop on introductory concepts in bayesian data analysis

Notebooks

The workshop comes in two slightly different versions:

University Workshop

Version presented at the IGGI Conference 2020

* Introduction  

* Bayesian Gears  
    - From counts to probability  
    - Bayesian updating  
    - Likelihood, Parameters, Prior and Posterior  
    
* Bayesian Machinery  
    - Parameters Estimation  
    - Grid Search, Quadratic Approximation, MCMC  

* Bayesian Models  
    - PyMC3 Model Building  
    - Linear Regression  
    - Logistic Regression  
    - Graphical Models  

Applied Workshop

Application-oriented version, more suitable for being delivered in industry settings

* Introduction

* Bayesian Approach to Inference
  - Counts
  - Updating Counts
  - From Counts to Probability
  - Likelihood, Parameters, Prior and Posterior
  - Parameters Estimation
  - Bayesian Models

* PyMC3
  - Model Building
  - Model Inspecting
  - Model Fitting
  - Model Evaluating and Comparing
  - Model Predicting

* Applications
  - PyMC3 vs scikit-learn
  - Web Traffic Estimation
  - Advertising Effect on Revenue
  - Game Difficulty Estimation
  - Model Comparison

Links

• Statistical Rethinking (most of the content has been adapted from here)
• PyMC3
• Papers, Please (images and narrative framework for the University Workshop version)

Requirements

1. Download your local version of the workshop repository
2. Install Anaconda
3. If you run a Windows machine and do not have administrator rights:

• Open Anaconda Navigator
• Create a new environment with python=3.6 and call it workshop_env
• Open the Anaconda Powershell Prompt associated to the new environment
• Navigate to the workshop directory

3. If you run a Windows machine and do have administrator rights:

• Open the Anaconda Powershell Prompt in the workshop directory
• Then from the Prompt:

# create anaconda environment
conda create -n workshop_env python=3.6

# activate the environment
conda activate workshop_env

4. At this point install all the requirements with:

# install the requirements
conda install -c conda-forge --file requirements.txt

# open jupyter 
jupyter notebook

5. Navigate to and open the .ipynb file of interest
6. Read the Usage section in the RISE documentation for navigating the notebook slides.