ABSTRACT
Multilevel regression with post-stratification (MRP) is a popular way to adjust non-representative surveys to analyze opinion and other responses. It uses a regression model to relate individual-level survey responses to various characteristics and then rebuilds the sample to better match the population. However, it can be a challenge to get started with MRP as the terminology may be unfamiliar, and the data requirements can be onerous. In general, MRP is a good way to accomplish specific aims, but it is not without trade-offs. The Nationscape dataset is a high-quality survey. But it was weighted to major census region—the West, the Midwest, the Northeast, and the South—rather than state, which could be one reason the people see a difference between the MRP estimates and the raw data. In a similar manner to Ghitza and Gelman (2020) pretend the people have access to a US voter file record from a private company.
