SAE 2022 - Small Area Estimation of (Non-Linear) Indicators Using Mixed Effects Random Forests

Abstract

Reliable estimators of the spatial distribution of socio-economic indicators are essential for evidence-based policymaking. As sample sizes are small for highly disaggregated domains, the accuracy of the direct estimates is reduced. To overcome this problem small area estimation approaches are promising. This presentation promotes the use of random forests as versatile tools for estimating spatially disaggregated indicators in the presence of small area-specific sample sizes. Mixed effects random forests (MERF) combine advantages of regression forests (robustness against outliers and implicit model-selection) with the ability to model hierarchical dependencies. In the presentation, I aim to provide a coherent framework based on MERFs for estimating various economic and poverty indicators in the presence of micro-data. Additionally, we developed two non-parametric bootstrap estimators for assessing the uncertainty of the estimates. Extensive model-based simulations show the advantages of the proposed point estimator and its uncertainty measure. Our proposed methodology is applied to income data from Mozambique.

Dr. Patrick Krennmair

Research Associate in Applied Statistics

I am working as a research associate at the Chair of Applied Statistics at Freie Universität Berlin and as a consultant for the statistical consulting unit fu:stat.