Online Program

Small area modelling under complex survey designs for business data
Jan Pablo Burgard, University of Trier 
*Ralf T. Munnich, University of Trier 
Thomas Zimmermann, University of Trier 

Keywords: small area estimation, complex surveys, business data, augmented modelling

Recent developments in small area statistics applications have raised the importance of considering the sampling design adequately in the model-based estimators. This can be achieved either by including survey weights into the estimation process or by using the design variables directly to augment the model.

The present study focuses on the interplay of modeling and survey weighting. The estimators of interest cover classical unit- and area-level small area estimators, weighted and augmented small area estimators and robust small area estimators. The designs considered comprise stratified and PiPS designs including the optimal and box-constraint optimal allocation. A design-based simulation study illustrates the impact of design-effects and the variability of design weights on model versus design-based estimation methods. This study is based on a close to reality dataset generated from the Italian ASIA and PMI data.

The work is conducted within the BLUE-ETS research project financed by the European Commission under FP7 (cf.