BAYESIAN CAUSAL INFERENCE

ABS Schools

The Applied Bayesian Statistics summer school has been running since 2004. It is organised by

• IMATI CNR Istituto di Matematica Applicata e Tecnologie Informatiche, Consiglio Nazionale delle Ricerche, Milano

Local contributing organization:
Florence Center for Data Science (FDS)
Department of Statistics, Computer Science, Applications (DISIA)
University of Florence

The school aims to present state-of-the-art Bayesian applications, inviting leading experts in their field.
Each year a different topic is chosen.

Past editions were devoted to: Gene Expression Genomics, Decision Modelling in Health Care, Spatial Data in Environmental and Health Sciences, Bayesian Methods and Econometrics, Bayesian Decision Problems in Biostatistics and Clinical Trials, Bayesian Methodology for Clustering, Classification and Categorical Data Analysis, Bayesian Machine Learning with Biomedical Applications, Hierarchical Modeling for Environmental Processes, Stochastic Modelling for Systems Biology, Bayesian Methods for Variable Selection with Applications to High-Dimensional Data and Applied Bayesian Nonparametrics, Modern Bayesian Methods and Computing for the Social Sciences, Bayes Big Data and the Internet, Modeling Spatial And Spatio-Temporal Data With Environmental Applications, Bayesian Statistical Modelling and Analysis in Sport, Bayesian Demography.

The lecturer is Prof. Fan Li (Department of Statistical Science, Duke University, Durham, NC, USA).
The practical lectures will be given by Veronica Ballerini, Fiammetta Menchetti and Giacomo Petrillo (DiSIA, University of Florence).

The topic chosen for the school is
Bayesian Causal Inference.

The aim of this course is to introduce the fundamental concepts and the state-of-the-art methods for causal inference under the potential outcomes framework, with an emphasis on the Bayesian inferential paradigm.
Topics will cover randomized experiments, common methods for observational studies, such as propensity score, matching, weighting and doubly-robust estimation, heterogeneous treatment effects, sensitivity analysis, instrumental variables, principal stratification, panel data methods, and longitudinal treatments. Recent advances related to high dimensional analysis and machine learning will be naturally incorporated into the discussion. All methods will be illustrated via real world case studies.

Tentative Program

• Day 1: Fundamentals of the potential outcomes framework. Randomized experiments and methods for observational studies, including propensity scores, regression adjustment, matching, weighting, doubly-robust. Case studies on debit card use and household spending.
• Day 2: General structure of Bayesian causal inference, model specification (linear, BART, GP, Bayesian casual forests, and other popular machine learning methods), heterogeneous treatment effects, implications in high dimensions, choice of prior.
• Day 3: Role of propensity scores in Bayesian causal inference and sensitivity analysis.
• Day 4: From instrumental variables to principal stratification, and implementation via STAN. Multiple case studies, including randomized experiment with noncompliance, and a regression discontinuity design.
• Day 5: Methods for panel data: difference-in-differences and synthetic controls. Methods on longitudinal treatment: g-computation and marginal structural models, and their Bayesian versions. Multiple case studies in political science and medicine. Concluding remarks.