Multilevel Regression and Post-Stratification

Abstract

Weighting Adjustments as Correctives for Survey Bias

Regression as Stratification

In this project I sought to understand the procedure of post-stratification adjustment used in election forecasting and regression modelling of the same. I published the documentation for the technique of using Multilevel Regression and post-stratification (MrP) with the bambi package here. In this work i tried to elaborate precisely how and why the careful modeller would want to make stratum specific adjustments to the predictions of regression models.

I adapted the work of Martin, Philips and Gelmen’s “Multilevel Regression and Poststratification Case Studies” that can be found here. You can download the worked example here

Comparison with OLS