Evaluation of Synthetic Small-area Estimators Using Design-based Methods

Authors

DOI:

https://doi.org/10.17713/ajs.v48i4.790

Abstract

The use of area-specific design-based mean squared error (MSE) to measure the uncertainty associated with synthetic and direct estimators is appealing since the same model-free criterion is applied. However, the small sample size is often a difficulty in obtaining a reliable estimator of the area-specific design-based MSE. Moreover, the area-specific design-based mean squared error estimator might yield undesirable negative values under certain circumstances. The existing solution to overcome the problem of small sample size is to consider average design-based MSE, average being taken over the available small areas. This may not solve the other problem of negative MSE. An alternative average design-based mean squared error estimator is proposed which always produces positive estimates. Simulation shows that this estimator performs better than the existing average design-based MSEs as it always produces positive estimates and accounts for the bias component usually present in synthetic estimators.

Author Biographies

Partha Lahiri, University of Maryland

Professor, Joint Program in Survey Methodology, University of Maryland, College Park, MD

Santanu Pramanik, Public Health Foundation of India

Research Scientist, Public Health Foundation of India, Gurgaon

Published

2019-07-25

How to Cite

Lahiri, P., & Pramanik, S. (2019). Evaluation of Synthetic Small-area Estimators Using Design-based Methods. Austrian Journal of Statistics, 48(4), 43-57. https://doi.org/10.17713/ajs.v48i4.790