Estimación de la tasa de desempleo e ingreso medio del departamento de Cundinamarca utilizando estimación en áreas pequeñas

Abstract

In this article, we propose using principal components analysis in a context of high-dimensional spaces to obtain orthogonal variables to use as auxiliary information in Fay-Herriot (FH) and spatial Fay-Herriot (FH). The accuracy are computed for direct, Fay-Herriot and proposed estimator. An empirical application with R ecosystem (dplyr, ggplot2, survey, and other R packages) is carried out using 2017 Bogotá Multipurpose Survey (EMB2017), a household survey with socioeconomic and demographic information of Bogotá and surrounding municipalities. In particular, percapita average household income and unemployment rate are estimated, then a benchmark of proposal estimators with classical FH estimator is carried out (Molina, 2015). The work illustrates how the spatial Fay - Herriot estimator with selected components from PCA has the best performance in terms of accuracy.