Small area estimation

In the dissertation special problems that may be encountered in finding optimal estimation strategy for small area estimation, in particular, model diagnostics for small area models, constrained estimation, sample design selection, nonresponse adjustment and borrowing strength across both small areas and time are considered.

The estimation strategy is a combination of sampling design and estimation design. First of all several well known sample designs and population total estimators are constructed for small areas. The changes of these estimators are showed in the case of the use of various nonresponse adjustment methods. A simulation using a real population from Statistics Lithuania is done to investigate the performance of different types of estimation strategies when various problems occurs: small area, nonresponse. The different underlying models are examined for both design-based model assisted and model-based estimators.

This study showed that the nonresponse has bigger negative effect for design-based estimators than for model-based estimators. Still, generally, the design-based model assisted estimator performs better than model-based estimator. To improve estimation strategy a balance sample and model-based sample design are introduced. The model-based sample design is based on the historical data, which are used to construct superpopulation model before sample selection. The variance of prediction error is used to construct inclusion probabilities, thus the element with larger variance of prediction error have larger probability to be selected in to the sample. The simulation studies showed, that in many cases the use of model-based sample design reduce the accuracy measures.

Read electronic version of the book:

DOI: https://doi.org/10.20334/2099-M