Outlier robust small domain estimation via bias correction and robust bootstrapping

Several methods have been devised to mitigate the effects of outlier values on survey estimates. If outliers are a concern for estimation of population quantities, it is even more necessary to pay attention to them in a small area estimation (SAE) context, where sample size is usually very small and the estimation in often model based. In this paper we set two goals: The first is to review recent developments in outlier robust SAE. In particular, we focus on the use of partial bias corrections when outlier robust fitted values under a working model generate biased predictions from sample data containing representative outliers. Then we propose an outlier robust bootstrap MSE estimator for M-quantile based small area predictors which considers a bounded-block-bootstrap approach. We illustrate these methods through model based and design based simulations and in the context of a particular survey data set that has many of the outlier characteristics that are observed in business surveys.