Outcome-Dependent Sampling for Continuous Longitudinal Response Data

• Activity Number: 139
• Type: Contributed
• Date/Time: Monday, August 2, 2010 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Jonathan Schildcrout*+ and Paul J. Rathouz and Patrick Heagerty
• Companies: Vanderbilt University and The University of Chicago and University of Washington
• Keywords: longitudinal data ; outcome dependent sampling ; epidemiological study designs

Abstract

Outcome dependent sampling (ODS) designs are crucial when prospective designs are infeasible or when ascertainment costs limit study sample size. In univariate settings they are ubiquitous. However, with longitudinal response data ODS designs are far less common because analyses are not straightforward, and little work has been done for designs involving continuous outcomes. For example, maximum likelihood analyses require ascertainment corrections which are not available in most software packages. We discuss a class of ODS design where sampling is based on features of the continuous longitudinal response vectors, and we propose an ascertainment corrected likelihood analysis. Methods are motivated by and will be applied to examination of genetic modification of lung growth in adolescents participating in the Childhood Asthma Management Program Continuation Study.