Statistics and Data Science II

Aim

After completion of the course, the student should at an advanced level be able to:

• use statistical software to estimate appropriate linear regression models for cross-sectional and panel data and explain the statistical principles underlying these estimates;
• use statistical software to assemble appropriate data structures for estimating regression models and implementing robustness checks;
• interpret the parameters of linear regression models, produce predictions, and evaluate goodness of fit;
• describe the logic of causal inference and how it applies to regression models, distinguishing between causality and correlation;
• identify common threats to causal interpretation of linear models, and assess and justify modeling approaches for solving these threats;
• evaluate the validity and robustness of causal inferences under a variety of assumptions about how the data was generated.

Content

This course introduces the principles and practice of linear regression modeling. Underlying model assumptions are reviewed and scrutinized. In intensive computer laboratories, statistical tools for creating appropriate data structures and estimating models using real data are presented and guidance is provided in interpretation of model parameters. The remainder of the course focuses on causal inference and the potential outcomes framework. Panel data models and statistical tools for their estimation are presented, and their potential to improve causal inference are compared. Discussion is extended to consider natural experiments and instrumental variable approaches to causal inference. The sensitivity of estimates to violations of model assumptions are evaluated, with special attention given to methods centering on computer simulation.