Bayesian Modeling
Instructors: Dr. Sascha Göbel
Abstract
Modern data science requires substantial flexibility in setting up complex statistical models, thrives from incorporating existing knowledge or beliefs into analyses, and demands intuitive quantification of uncertainty. This is especially relevant when dealing with hierarchically structured data, constructs that elude direct measurement, and settings rich in historical and contextual information, as typical for political and policy-relevant data. Here, adequate modeling tools and approaches are essential. This course provides a foundational and practical introduction to Bayesian hierarchical, i.e., multilevel, and latent variable modeling, which adheres to this demand. Students will engage with the foundations of the Bayesian approach, understand the inherently Bayesian nature as well as the utility of multilevel and latent variable models, learn how to program, fit, and evaluate such models using the modern probabilistic programming language Stan, get a glimpse at the underlying machinery, and practice applying this knowledge in the context of their own and contemporary research questions. This is an advanced course intended for students with prior knowledge of statistics at the level of Statistics II: Statistical Modeling and Causal Inference and prior knowledge of working with R at the level of Introduction to Data Science.
Find out more