Using MRP to generalize from sample to population: A step by step example

Location image of event venue


Multilevel regression and poststratification is becoming increasingly popular as an adjustment method for survey nonresponse and non-representative data more generally. At this meetup we will work through how to implement MRP using a toy example in both Rstanarm and (if we have time) Stan. As we work through the example we will highlight what makes MRP so useful and some of the necessary conditions that MRP needs to be suitable. This meetup will be relatively practical with R code provided. Attendees should feel welcome to bring laptops to follow along (requirements might include R, tidyverse, rstanarm, bayesplot, shinystan and rstan), but this is not necessary. This is joint work with Andrew Gelman and Jonah Gabry.

Lauren is a postdoctoral research scientist in a joint position between Columbia Population Research Center and the Department of Statistics, Columbia University. She obtained her PhD in psychology at the University of Adelaide. Lauren develops survey weights and methodology for the New York City Longitudinal Study of Wellbeing.