Thomas Lumley • 3/9/2024

Small-area estimates by smoothing direct estimates

The article discusses small-area estimation techniques in survey statistics, where direct estimators for domains are too variable. It explains how to borrow information across domains using weighted averages and linear mixed models, specifically the Fay-Herriot model, to reduce variance at the cost of bias. It also covers spatial smoothing for geographical areas and Bayesian approaches, noting challenges in applying these methods to non-geographical domains.

0 comments

#Statistical Modeling #Linear Mixed Models #Survey Estimation