J. P. Burgard, M. D. Esteban Lefler, D. Morales, R. Münnich, A. Perez Martin
In small area estimation, when using area level models, it is assumed that the covariates are measured precisely. The covariates do not induce any additional sources of variability into the model. However, in practice, this assumption is violated frequently. The lack of available exact information, e.g., from registers, forces researchers to use estimated covariates in their models. By using random variables as covariates, the structure of the model used for the estimation changes considerably. When ignoring this source of variability the model parameters are biased estimated and both, the point and MSE estimation, decay in their precision. We propose a model that accounts for the variability in the estimated covariates, provide well functioning estimation procedures and an analytical MSE approximation. We show the superior performance of our proposed estimator by simulations and on a real life application.
Palabras clave / Keywords: small area estimation, measurement error, MSE estimation, area level model
Programado
Sesión bilateral SEIO-DStatG: Small Area Estimation (Organizadores: Ralf Münnich y Domingo Morales)
1 de junio de 2018 12:30
Sala Cristal
