Outlier Robust Small Domain Estimation for Business Surveys in the Netherlands

The Netherlands Central Bureau of Statistics recently investigated the use of small domain estimation methods in business surveys with emphasis on outlier robust methodologies. In this talk we discuss some of the approaches we considered for this purpose. In particular, we present outlier robust domain estimation using random effects and M-quantile models as well as robust regression synthetic and GREG-type estimators. Attention is paid to the use of design weights, their role in protecting against model misspecification and their relation to influence weights. The alternative methodologies are applied to data from one industry sector of the Dutch Structural Business Survey.