Publications by authors named "Swen Kuh"
Article Synopsis
- Multilevel regression and poststratification (MRP) is increasingly used for population inference, but its accuracy can be influenced by model details and validation practices are under-researched.
- The study investigates leave-one-out cross-validation (LOO) methods, focusing on Pareto smoothed importance sampling (PSIS-LOO) and a survey-weighted alternative (WTD-PSIS-LOO), using simulations to assess their effectiveness in ranking model performance.
- Findings reveal that while both PSIS-LOO and WTD-PSIS-LOO can identify the best and worst models, they struggle with accurately ranking models for MRP, particularly in small-area estimations, indicating a need for caution in relying solely
