This is an introductory course in Bayesian statistical modelling. We will read chapters of the textbook by McElreath (2020, 2nd edition), discuss the content and apply the methods in exercises using the brms package in R.
Learning objectives
Students ...
… have experienced and understood the fundamental philosophy behind Bayesian probability theory,
… have acquired the skills to do Bayesian analysis using the brms package in R,
… know which resources to consult for further study.
Topics
- Philosophical difference between classical and Bayesian statistics
- The R package ‘brms’
- Numerics: grid approximation; quadratic approximation; Markov Chain Monte Carlo
- Working with samples from posterior; posterior predictive checks; prior predictive simulation
- Linear regression
- Categorical predictors; interactions
- Confounding effects; model comparison; regularizing priors
- Generalised Linear Models: Binomial regression; Poisson regression; over-dispersion; zero-inflation
- Hierarchical models: Varying intercepts; varying slopes; multi-level posterior prediction choices
Format
The mode of working is a mix of independent textbook study; collective discussion; independent and collective problem solving; homework; and lecture-style inputs from the teacher as needed.
The open-source software STAN will be used via the brms package R. An introduction to and help with brms/STAN will be provided. Students need a good working knowledge of R!
Homework will be submitted using R Markdown.
Allocation of places
Due to the mode of working in this course places are limited. Students are required to register via Agnes. Priority will be given to 4th semester students of the Global Change Geography Master. |
