Analysis of Biased Sampling in Longitudinal Data

• Activity Number: 483
• Type: Topic Contributed
• Date/Time: Wednesday, August 12, 2015 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Leila Zelnick* and Patrick Heagerty and Jonathan Schildcrout
• Companies: University of Washington and University of Washington and Vanderbilt University
• Keywords: experimental design ; epidemiology ; biased sampling

Abstract

Investigations commonly use data that were originally collected for a different primary research goal. When novel biomarkers are investigated, it is natural to leverage existing data together with stored specimens to yield new exposure data. Limited availability of specimens and financial resources may require targeting a subset of patients for new analyses; use of outcome dependent sampling (ODS) designs can provide a cost-effective and efficient way to conduct substudies (Schildcrout et al. (2013)). To date statistical methods have focused exclusively on using only data from the subsample for final analysis, but research in the univariate setting (Weaver and Zhou (2005)) suggests that analyzing unsubsampled individuals, in addition to those on whom the biomarker has been ascertained, may contribute to improved estimation of target regression parameters. This talk will focus on the use of ODS sampling designs and analysis in the longitudinal setting with continuous outcomes. We examine the potential advantages of these designs/analyses from a likelihood perspective, and propose some alternatives to this approach. Results will be illustrated using a longitudinal cohort study.