A Balanced Sampling Approach for Multi-way Stratification Designs for Small Area Estimation

The goal of this paper is to analyse the possibility to improve the performances of the estimation at sub-regional levels for Istat surveys. In particular, we refer to small domains cutting across survey strata; e.g. estima-tion of unemployment rate for Local Labour Market Areas- defined as aggrega-tion of municipalities – currently estimated by means of EBLUP based on linear mixed model with spatially correlated area effects. In this work we exploit the use of semi-parametric models based on penalized splines to prevent misspecifications of the functional relationship between the target variable and some covariates. Moreover, in the same way, semi-paramet-ric modelling can be used to include more accurate spatial information using low-rank thin plate splines (Opsomer et al, 2008). Finally, we analyse the possibility of reducing bias of EBLUP through the introduction of model based direct estimators (Chambers e Chandra, 2006). The performances of the suggested methods are reported on simulated experi-ments on 2001 Census data. References Chambers, R.L. and Chandra, H. (2006). Improved Direct Estimators for Small Areas. Submitted fo