Causal Inference under the Rubin Causal Model: Experiments, Observational Studies, and Everything In Between
Room | Time
Description
A key requirement for social, scientific, medical, and technological progress is the ability to draw causal conclusions from data. A principled and powerful paradigm to address the task of causal inference is the Rubin Causal Model. This framework is founded on the concept of potential outcomes, which were effectively first formalized for randomized experiments by Jerzy Neyman and then extended to observational studies (and everything between controlled experiments and uncontrolled studies) by Donald Rubin.
This workshop will introduce students to frequentist and Bayesian statistical methodologies for drawing causal inferences from randomized experiments and observational studies under the modern Rubin Causal Model. Topics include the fundamental components of the Rubin Causal Model (Science, Learning, Decisions), causal inferences for classical randomized experiments, drawing causal inferences from linear regression models, the design and analysis of observational studies (propensity score subclassification and matching), and design and machine learning for Big Observational Data. The methodologies taught in this workshop possess a broad scope of application, ranging from the physical, life, social, and management sciences, among other domains. Conceptual understanding, rather than theoretical derivations of equations, will be emphasized throughout. Students will have a hands-on experience working with data using R and Stan in a Jupyter Notebook environment.