Olivia Ståhl: Estimating a population total when the distribution is highly skewed to the right

In the case of highly skewed populations the design-based Horvitz-Thompson estimator will tend to have a large variance. Model-based estimators are able to accommodate external information about the population shape into the estimation procedure, but instead introduce bias. As a way to address this trade-off we propose an estimator that combines design- and model-based properties; the estimator can be used to improve precision when the population is known to be skewed to the right but no auxiliary data is available.

In my talk I will start by introducing the most basic forms of design-based and model-based estimators and shortly discuss their properties. After that the combined estimator will be introduced and some preliminary results presented.