Title: A Simple Adaptation of Variable Selection Software for Regression Models to Select Variables in Nested Error Regression Models

Issue: Volume 81 Series B Part 2 Year 2019
Pages: 302 -- 317
Data users often apply standard regression model selection criteria to select variables in nested error regression models, which are widely used in small area estimation. We demonstrate through a Monte Carlo simulation study that this practice may lead to selection of a non-optimal or incorrect model. To assist data users who wish to use standard regression software, we propose a transformation of the data so that transformed data follow a standard regression model. Thus, variable selection software available for the standard regression model can be directly applied to the transformed data. We illustrate our methodology using survey and satellite data for corn and soybeans in 12 Iowa counties.
Primary 62J05; Secondary 62F07.
Keywords and phrases: Fuller-Battese transformation, Intracluster correlation, Lahiri-Li transformation, Variable selection criteria