PoL workshop on statistical inference
Physical biology is a quantitative science, and biological scientists need to be equipped with tools to analyze quantitative data. Bayesian methods are especially apt in this context because they involve explicit models that describe the data generation process. The Bayesian approach allows us to directly learn from experiment to inform our understanding of nature through models.
Prerequisites
Familiarity with the Python programming language and the NumPy package and Pandas or Polars packages are assumed. We will use Bokeh for plotting, and some familiarity with that package is useful as well, but not required. Finally, students should have some experience with calculus and familiarity with basic concepts in probability, though advanced knowledge is not required.
Data sets
Students should download and unzip the data sets we will use in the workshop.
Instructor
Lessons
- Lesson 0: Setting up computing resources
- Lesson 1: Introduction to Bayesian modeling
- Lesson 2: Parameter estimation by optimization
- Lesson 3: Markov chain Monte Carlo and Stan
- Lesson 4: Display of MCMC results and diagnostics
- Lesson 5: Prior and posterior predictive checks
- Lesson 6: Hierarchical models
- Lesson 7: Principled pipelines
Exercises
- Exercise 1. Practice with Bayesian modeling
- Exercise 2. Paremeter estimation by optimization
- Exercise 3. First foray into MCMC
- Exercise 4: More Bayesian inference with MCMC
- Exercise 5: Bayesian modeling with prior and posterior predictive checks
- Exercise 6: Hierarchical models
- Exercise 7: Principled pipelines
Schedule