POLS 208 Lecture Notes - Lecture 14: Terry Mcauliffe, Tariqa, Confounding
5 Dec 2018
School
Department
Course
Professor
Document Summary
Measuring if an armband actually improves your # of free throws. Someone who is jeff in a parallel universe but without an armband, who we can compare the end result of # of free throws as comparison. We still have to believe anthony is a good counterfactual for jeff. Someone with the same starting point to observe the effect because if the slope is the same. Because otherwise, jeff was improving at the same rate as anthony (same slope) Diff in diff estimate = (difference in treatment) - (difference in control) Suppose the arm band didn"t matter - the treatment effect would be 0 and the slope would be the same. Outcome trend did not determine allocation of treatment. Amazon doesn"t deliver cool new arm bands to players whose growth rate is faster. Ordering an arm band in the first place is not related to growth rate (is this problematic?)
