Title: Statistical Modeling in Survey Sampling: Some Solutions to Ever Increasing Demand for Reliable Data
Speaker: Gauri S. Datta, University of Georgia and US Census Bureau
Background:
The modern society shows an ever-increasing appetite for reliable and up to date data to make informed decisions in both public and private sectors alike. While censuses, usually conducted once in a decade, provide reliable information about the population across various geography and demography, such information quickly get outdated each passing year after the census. To obtain a current picture of the population under study, sample surveys are frequently conducted by various agencies to collect data from only a fraction of the population. Due to budget constraints, these surveys are inherently limited in size. While information gained from such surveys may be adequate for the entire population, the same data is often inadequately small when it is sliced and diced across geographic and demographic sub-populations. These sub-populations are termed small areas.
Statistical summaries based on traditional direct estimates, depending only on sample data from individual small areas, are usually very unreliable. In Small Area Estimation, appropriate and useful statistical methodologies have been developed to improve on the traditional direct estimates. By “borrowing strength” from other data sources, improvement over the direct estimates is sought by developing reliable statistical methods by using suitable models. A course on small area estimation will deal with this body of statistical methodology and its applications to some important inference problems, faced by government agencies or private sectors.
Course Content:
This short course will be based on the book “Small Area Estimation, by J.N.K. Rao and Isabel Molina, 2015, Wiley”, and a number of research papers written by researchers on the topic. Topics include introduction to small area estimation, direct and indirect estimators in domain estimation, model-based approaches to small area estimation, area-level and unit-level models, empirical best linear unbiased prediction for point estimation and corresponding mean squared error estimation, small area estimation applications by R, hierarchical Bayes and empirical Bayes methods in small area estimation, and more applications of R.
Presentation Plan:
- Introduction to small area estimation: Direct and Indirect estimators in domain estimation.
- Mixed effects models for small area estimation: Area-level and unit-level models
- Empirical best linear unbiased prediction: point estimation and corresponding mean squared error estimation
- Small Area Estimation applications using R
- Hierarchical Bayes and empirical Bayes methods in small area estimation. More applications using R