There is a growing demand to produce reliable estimates of various socio-economic and health characteristics at both national and sub-national levels. However, data availability at the sub-national (small area) level from a survey is often limited by cost and thus analysts must make the best possible use of all available information. The course will begin with a history of small-area estimation and different uses of small-area statistics in both public and private sectors. This course will provide an introduction to the main concepts and issues in small estimation and describes various approaches for estimating different small area parameters. Topics include standard design-based methods, various traditional indirect methods and the state-of-the-art small-area estimation methods that use both Bayesian and empirical best prediction methods. Monte Carlo simulation results and data analysis using available statistical software will be presented.
a. Uses of small area statistics.
b. Different data sources for producing small area estimates.
c. A few real life applications.
2. Traditional Indirect Methods
a. Synthetic methods
b. Composite Methods
3. Model-based methods
a. Relevance of mixed models in small area estimation.
b. Area specific versus unit specific mixed models.
4. Implementation of a mixed model
a. Empirical best prediction (EBP) method.
b. Hierarchical Bayes method.
5. Case Studies
Grading will be based on:
The course is intended for survey practitioners and should be accessible to graduate students and early career researchers. An undergraduate level course in mathematical statistics and applied regression analysis are required. If you are unsure about your qualifications for the course, please contact us.