POLS 208 Lecture Notes - Lecture 13: Data Collection, Natural Experiment, Medicaid
5 Dec 2018
School
Department
Course
Professor
Document Summary
The asterisk shows that its statistically significant. Nothing in the model distinguishes your main independent variable or treatment and the control variables. When you write a regression model, you are assuming that there is no bias from omitted variables in your models, that you know exactly how things came about. Be aware of assumptions implicit in regression analysis. The answer will ultimately rely on the credibility of the assumptions. One way to think of omitted variable bias. Foreign is negatively correlated to weight and positively correlated with price. If there"s an unobserved confounder that we suspect is positively related to both treatment and outcome, we will overestimate the effect of treatment. If the confounder is negatively related to both treatment and the outcome again, we will overestimate the effect of treatment. If the confounder is positively related to treatment but negatively related to outcome, or vice versa, we will understate the effect of treatment.
