Estimation of small area means using area-level and unit-level covariates based on multiple surveys


Small Area Estimation with Data from Multiple Sources

Gauri Datta (University of Georgia, U.S. Census Bureau, and RTI International) (Speaker)
J. N. K. Rao (Carleton University)
Mahmoud Torabi (University of Manitoba, MB, Canada)
Benmei Liu (Division of Cancer Control and Population Sciences, National Cancer Institute, MD, USA)

Unit-level models are extensively used in small area estimation. These models incorporate both unit-level and area-level covariates to accurately estimate finite population means of small areas. To borrow information from the unit-level covariates, that are available only from the sampled units, we propose a multivariate adaptation of the nested error regression model. Information on the area-level covariates are gathered from separate surveys and are incorporated into our model via measurement error modeling of covariates. We develop empirical best linear unbiased predictions of small area means and associated mean squared error of prediction. Performance of the proposed approach is studied through simulation studies and also by a real application. We apply our methodology to estimate mean BMIs of several demographic groups based on data from the U.S. National Health and Nutrition Examination Survey and the National Health Interview Survey.