D. Morales González, A. Perez Martin, L. Santamaria Arana

Statistical agencies produce parameter estimates for domains of different levels of aggregation. They are interested to report estimates of linear parameters of domains and subdomains in a consistent way; this is to say, with the property that the sum of the subdomains estimates match up with the corresponding estimate of its domain. The two-fold nested error regression model provides EBLUP estimates of linear parameters of domains and subdomains with this desired property. Three methods to fit the model and a closed-formula procedure to estimate the mean squared error of the EBLUP estimates are given and empirically studied. Results of three simulations studies are reported. An application to Spanish Living Conditions Survey data is given.

Palabras clave: small area estimation linear mixed models EBLUP mean squared error living condition survey

E. López Vizcaíno, M. J. Lombardía Cortiña, D. Morales González

A. Fernández Militino, L. Ugarte Martínez, T. Goicoa