3.3: Model Specification Strategies - Class Notes

Contents

Tuesday, October 29, 2019

Overview

Today we will simply talk about the process of specifying a model: how to pick X variables to add to your regression. Recall, they should fit the two criteria for omitted variable bias. One additional issue we discuss is the use of proxy variables: where you know omitting a variable would cause bias, but you don’t have data on it, just data on a variable related to it; and of control variables (including proxies), where the OLS estimator (^βj) for an independent variable is not causal, but just helps remove bias on the independent variables you do care about being causal.

For the rest of class, we will work on practice problems for multivariate regression.

Slides

• Lecture Slides

Practice Problems

Today you will be working on R practice problems. Check back later for solutions.