Analysis of Biased Sampling Designs in Longitudinal Data

• Activity Number: 126
• Type: Contributed
• Date/Time: Monday, August 4, 2014 : 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: longitudinal data analysis ; epidemiological study design ; biased sampling

Abstract

The use of outcome dependent sampling in longitudinal data analysis has been shown to improve efficiency in estimating regression parameters, when outcome data exist for all cohort members but key exposure variables will be gathered only on a subset (Schildcrout and Heagerty, 2013; Neuhaus, Scott, and Wild, 2006, among others). Inference incorporating the information from individuals whose exposure has not been ascertained has previously been investigated for univariate but not longitudinal outcomes (Weaver and Zhou, 2005). For a continuous outcome and a simple binary covariate, we explore the contributions of various sources of information to the estimation of key regression parameters in a likelihood framework. Furthermore, we evaluate the efficiency gains that the associated estimators might offer over random sampling and offer insight into their relative merits in a few key scenarios.