Mathematics of Information Technology and Complex Systems


Project Homepage
 
Project Highlights
 
Research
 
Team Members
 
Partner Organizations
 
Students
 
Internship Positions
 
Publications
 
Presentations
 
Events
 
MITACS Home
 

Statistical Methods for Complex Survey Data


Project Leader: Dr. Changbao Wu , University of Waterloo


Survey data now being collected by many government, private, health and social science organizations have increasingly complex structures precipitating an urgent demand for new statistical methodology to further research in substantive areas. In cross-sectional studies, which are taken at one point in time, it is typical to use very complex sampling designs, involving stratification and clustering as the components of random sampling. There can also be complexities in the resulting data file due to the patterns of nonresponse. In longitudinal studies, which follow individuals or groups of individuals over time, there is additional complexity stemming from possible complex correlation structures induced by repeated measurements on the same sampling unit, by irregularly spaced data and differing numbers of repeated observations on individuals. This data type, with all its various complexities, is increasingly common in substantive areas due to its power to infer causality, to separate individual and population trends and to detect changes in time.

Canada is a world leader in sample survey methodology and many of Canada's top researchers in this area are on this team. So too are some of Canada's top methodologists in longitudinal data analysis and hierarchical models, and in addition, are U.S. based researchers from Westat Inc. This project involves four overlapping groups of researchers with common research interests in the various complexities of surveys. Within these groups are researchers from our non-academic partners in government, private industry and biomedicine where this data type is created and utilized. Foremost among these is Statistics Canada, widely regarded as one of the leading national statistical agencies in the world, noted for the high quality of its data and its research and analysis. Statistics Canada has identified a pressing need for new methodologies as its responds to the demands for a wide variety of survey initiatives. The second group is from Westat Inc., a large U.S. survey methodology company in Washington, DC. that designs, collects, analyzes and researches new methodologies for U.S. government agencies, international government agencies as well as the private sector. Researchers from Westat involved in this project are interested in many aspects of complex survey data, but are primarily focused on research in variance estimation, replication-based methods, and issues of data disclosure. The third group is comprised of researchers in a number of subject matter disciplines. These researchers are creating the demand for these new and more complex surveys, and hence new methodologies. Increasingly, survey data are used to supplement existing databases in biomedicine, which has lead to the involvement of such partners with a common research focus. Researchers at the Toronto Rehabilitation Institute and the Institute for Clinical and Evaluative Sciences study the variability of medical outcomes and interventions across institutions and jurisdictions. It is of interest to them to incorporate related data housed in Statistics Canada.