CMStatistics 2022: Start Registration
View Submission - CMStatistics
B1782
Title: Challenges and successes with structured prior modelling in survey adjustment Authors:  Lauren Kennedy - Monash University (Australia) [presenting]
Abstract: Multilevel regression and poststratification have grown in popularity as a method to adjust for non-response and non-probability samples in surveys. One of the hallmarks is the use of regularization or partial pooling to ensure efficient predictions for rarer groups in the sample (either through sampling or through underlying population demographics). We will discuss how structured priors help to achieve this regularization, when they are useful to consider, and some of the complexities of identifying a ``good'' prior to use.