Resumen de An evaluation of the BLUP estimator in the practice of the small area estimation

Alex Costa, Eva Ventura, Albert Satorra Brucart

• A fundamental problem in small area estimation is that the BLUP of the parameter of interest (the small area parameter) is obtained as a linear combination of direct and a synthetic estimator which weights that have to be estimated from sample data. The result feasible BLUP is no more a linear estimator of the direct and synthetic estimator and it can even deviate substantially from the optimal predictor. Deviations from optimality can arise due to violation of distributional assumptions, at both the within and between area levels, heteroskedasticity on the within area variances, as well as not enough sample size both at the within and between area levels. In this paper we explore using Monte Carlo methods the quality of several feasible alternative small area estimators.

  Several concrete issues of small area estimation arising in IDESCAT (Statistics Bureau of Catalonia) have motivated our study. We use Monte Carlo methods in two scenarios: theoretical distributions and a specific population of labor statistics from the affiliation of firms in the INSS (National Institute of Social Security) registers.