Sample Survey Methods and Software for Non-survey Data Applications

• Activity Number: 391
• Type: Invited
• Date/Time: Thursday, August 12, 2004 : 8:30 AM to 10:20 AM
• Sponsor: Section on Survey Research Methods
• Author(s): Gary G. Koch*+ and Lisa M. LaVange
• Companies: University of North Carolina, Chapel Hill and Inspire Pharmaceuticals, Inc.
• Keywords: sample survey design ; clustering ; estimating equations ; randomization ; intraclass correlation ; multiple regression

Abstract

Survey data analysis approaches developed to accommodate clustering and unequal weighting in the sample design have more general application to clinical studies in which repeated measures structures are encountered, including multivisit studies, multiperiod cross-over studies, multicenter studies, and epidemiological studies of repeated illnesses. Multiple regression methods that allow for the specification of a primary sampling unit within which the repeated measures have intraclass correlation are available for logistic models, proportional odds models, and linear models. Survey methods for the analysis of clustered time-to-event measurements are applicable to clinical trials of survival data. More recently developed methods for nonparametric ANCOVA, motivated by sample survey techniques, have utility in randomized clinical trials, particularly in the presence of clustered randomization. A brief review of methods and software is provided, followed by examples from clinical studies. Areas of future research as well as the need for further software development in order to make these procedures more widely available to the nonsurvey audience will also be discussed.