Flexible Domain Prediction using Mixed Effects Random Forests

This paper promotes the use of random forests as versatile tools for estimating spatially disaggregated indicators in presence small area-specific sample sizes. Small area estimators are predominantly conceptualized within regression-setting and rely on linear mixed models to account hierarchical structure survey data. In contrast, machine learning methods offer non-linear non-parametric alternatives, combining excellent predictive performance a reduced risk model-misspecification. Mixed effects combine advantages regression with ability model dependencies. provides coherent framework based averages proposes bootstrap estimator assessing uncertainty estimates. We illustrate our proposed methodology using Mexican income-data from state Nuevo Le\'on. Finally, is evaluated model-based design-based simulations comparing traditional regression-based approaches averages.