Probabilistic models describe how the observed data was generated, and what structure the signal and noise from potentially multiple sources may have. Many classical statistical models are special cases of probabilistic models with special modeling assumptions. Probabilistic models can be implemented, improved, and critizised in a flexible, explicit and transparent manner, and the analysis can be supported with prior information about the data.
This 1-day course provides an introduction to Bayesian/probabilistic models. We will implement standard linear models based on the rstanarm package of the R statistical programming environment and readily available example data sets. The workshop is an ideal opportunity to familiarize yourself with the basic ideas in probabilistic modeling such as prior information, likelihood, model criticism and validation, as well as some of the available tools. At the end, you should be able to implement basic probabilistic models yourself, and understand their relative advantages and pitfalls compared to their classical alternatives.