This is the program for the 2010 Joint Statistical Meetings in Vancouver, British Columbia.
|CE_20C||Tue, 8/3/2010, 8:30 AM - 5:00 PM||CC-11(East)|
|Modeling and Data Analysis for Complex Surveys — Continuing Education Course|
|Instructor(s): Jean Opsomer, Colorado State University, Jay Breidt, Colorado State University|
|Survey data often arise from the efforts of institutions (like government agencies) to describe characteristics of a heterogeneous population in a cost-effective way. Typically, institutions use complex surveys with stratification, clustering, and unequal probabilities, and use various weighting adjustments to account for nonresponse and calibrate to external data sources. Researchers working with complex survey data may be tempted to ignore these sampling aspects when conducting statistical analyses, but traditional methods often are not appropriate unless the complexities of the survey design are explicitly taken into account. This short course is aimed at researchers with a basic background in statistical theory and methods (e.g. Neter, Kutner, Nachtsheim and Wasserman (1996), "Applied Linear Statistical Models") who need to analyze complex survey data. No previous background in survey sampling is assumed. We begin by reviewing the features that make survey data complex, including design properties and post-sampling adjustments, and a discussion of the potential pitfalls of ignoring these aspects during analysis. We then describe and compare several model- and design-based approaches to estimation and inference with complex survey data, highlighting their respective advantages and disadvantages. We demonstrate the various approaches with example data sets, including step-by-step illustrations using statistical software.|