POL222H1 Lecture Notes - Lecture 7: Propensity Score Matching, Instrumental Variable, Confounding
Document Summary
Pol222h alternative methods and validity of observational research. We want to fix the value of z control for z - Divide our data by different values of x and z. Then we use the conditional average of y and compare them across different values of x. How y would vary on average as x changes holding z constant. Alternative method to control for confounding variables when x is a dichotomous variable (when x is either 1 or 0) We say observation in a treatment group if x = 1 or a control group if x = 0. Look for matching observations between the two treatment groups (observations in the control group are matched with the observations in the treatment group (0 and 1) How y varies on average as x changes holding z constant. Instead of matching observations based on the values of each z, we may match them based on a summary of all values (check slide for rest)