Course Summary
This course is a continuation of Econ 626 and taught by professor Vadim Marmer . There is a discussion of linear econometric models, and their identification and estimation. Nonlinear models are covered in the context of extremum estimation. The general theory of extremum estimators is also studied with the main focus on generic approaches to establishing their consistency and asymptotic normality. The course concludes with topics in model (variables) selection using machine learning (ML) LASSO-based approaches. While the majority of the ML literature is concerned with prediction, the focus is on estimation of structural effects.
TA notes (preliminary and under continuous updating)
- Vector Differentiation Review
- Linear Algebra Review
- Linear Regression Review
- Misspecified IV as Bad Controls Problem
- Instrumental Variables Model Review
- Generalized Method of Moments Review
- Stochastic Equicontinuity
- Extremum Estimation Review
- Uniform Law of Large Numbers for Extremum Estimators
- Quantile Regression Review
- Bayesian Information Criterion Review
- Mixing Processes and Time Series
- Large Sample Theory with Time Series and Cluster Dependence
Useful Readings
- Davidson (1994) – This is an amazing textbook at the graduate level, which covers key concepts to understand large sample theory.
- Hansen (2020) – This paper applies the tools from time series theory to the case where there is cluster dependence. You will notice that using similar covariance inequality tools we can reach a central limit theorem with cluster dependence.