Small Area Estimation (Course code 0421)

Purpose

To familiarize participants with traditional and model-based methods for small area estimation and their application.

Benefits to Participant

Participants will gain an understanding of the methods for small area estimation, a topic of practical and theoretical interest due to growing demands for reliable small area statistics. Applications will demonstrate the implementation of the methods in practice.

Target Population

Employees who develop and implement sampling designs and estimation in the course of their work and wish to gain a thorough understanding of the methodology for small area estimation.

Course Outline

• Terminology, Examples
• Review of direct domain estimation
• Traditional indirect estimators: Synthetic, composite, James-Stein
• Small Area Models: area level and unit level models
• Linear mixed models and best linear unbiased prediction, major applications of basic area level and unit level models, model diagnostics, extensions
• Empirical Bayes estimation: confidence intervals, binary and count data, disease mapping applications, triple-goal estimation
• Hierarchical Bayes estimation: Markov Chain Monte Carlo methods, practical issues, applications

Prerequisite

Knowledge of mathematical statistics and exposure to a sampling theory course at an intermediate or advanced level. A course in linear regression and mixed models would be helpful in understanding model-based estimation.

TEXT BOOK: "Small Area Estimation" by J.N.K. Rao, Wiley, 2003.

Duration: 5 days