Linear models in small area estimation

Abstract

The possibility of obtaining subregional estimates of different crops on the basis of a regional sample currently employed in a survey on the characteristics of Italian farms is explored. Four linear models are rested, which employ the auxiliary information coming from the individual data of the Census of Agriculture at different extents. In general, a reduction of variability, expressed by the coefficient of variation, is obtained, but the strategy to adopt is not unique: there is to consider jointly the detail of modelization, i.e. the quantity of auxiliary information to employ, the size of the subareas together with the diffusion and incidence of the particular crop.