CFE-CMStatistics 2026: Registration
View Submission - CFECMStatistics2026
A1867
Title: From start to finish: Small area learning Authors:  Katarzyna Reluga - Humboldt University of Berlin (Germany) [presenting]
Nicola Salvati - University of Pisa (Italy)
Nikos Tzavidis - University of Southampton (United Kingdom)
Setareh Ranjbar - Lausanne University Hospital , University of Lausanne (Switzerland)
Abstract: Survey sampling methods are a cornerstone of statistical methodology. Small area estimation (SAE) improves on traditional direct estimators by borrowing information across domains, thereby increasing the effective sample size for areas with limited data. A framework for integrating machine learning with SAE is presented, an alliance complicated by the data scarcity inherent in small-area applications. A class of learners suited to different levels of data availability is introduced, together with a general procedure for selecting among them. Empirical comparisons demonstrate that the proposed methods outperform classical alternatives.