Prerequisites: 

STAT 420, SURV 440, or equivalents.

Description: 

This course covers the principles of model-based sampling, addressing such fundamental issues as the choice of working model, preferred estimation procedures, desirable sample designs, and protection against the model's being wrong. Model-based properties are studied of standard sampling designs such as simple random sampling, stratified random sampling, and multistage (cluster) sampling. Emphasis is on protection against bias and on robust variance estimation. Topics include: role of balanced samples in bias protection and optimality; relationship to balanced sampling of systematic sampling and probability proportional to size sampling; stratification and the use of models to guide sample allocation; estimation using samples from clustered populations; variance estimation in unclustered and clustered populations; incorporating quantitative and qualitative auxiliary data in estimating totals; comparison to design-based procedures like the general regression estimator. An important part of the course will be learning to program simulation studies in the R language. Students will be assigned small simulation problems as homeworks and a larger simulation project with a technical report.