Semiparametric Approaches for Auxiliary and Incomplete Covariate under Outcome Dependent Sampling Designs Public Deposited

Downloadable Content

Download PDF
Last Modified
  • March 19, 2019
  • Choi, Wansuk
    • Affiliation: Gillings School of Global Public Health, Department of Biostatistics
  • In epidemiologic and biological studies, investigators seek to establish relationships between a response variable and expensive risk factors. In reality, to obtain exact measurement about the covariate of interest can be difficult due to the limitation of budget or missingness of covariate while outcome and auxiliary information for covariate of interest are relatively cheap to collect. Under these circumstances, Outcome-Dependent sampling (ODS) and Outcome-Auxiliary dependent sampling (OADS) can be used to gain efficiency since those sampling designs incorporate additional information into parameter estimates. In this proposal, we propose three topics: (1) an estimated likelihood under ODS including missing in a covariate; (2) an updating method under a two-stage ODS for continuous response outcome; (3) an updating method under two-stage OADS for binary outcome. (1) and (2) are developed for continuous outcome variable case and (3) is developed to handle binary outcome variable cases. The first topic uses an estimated likelihood approach which is an extension of the method in Weaver and Zhou (2005) to a single-stage ODS sample that includes missing in a covariate of interest while the method in Weaver and Zhou (2005) is developed for two-stage cohort study without an auxiliary information. The second topic considers a semiparametric empirical likelihood method (Owen,1988, 1990; Qin and Lawless, 1994; Zhou et al., 2002, Wang and Zhou, 2006) at the second stage and updates estimators from the second stage by incorporating auxiliary information from data at the first stage. In the third topic we develop a new sampling scheme using auxiliary covariate information in a two-stage prospective study. For all three topics, the consistency and asymptotic distributions of proposed estimators are established, the finite sample performances are demonstrated through simulations studies, and real data application to the Collaborative Perinatal Project (CPP, Niswander and Gordon, 1972) are illustrated. The results from our methods show that one could gain efficiency by using auxiliary information.
Date of publication
Resource type
Rights statement
  • In Copyright
  • Becker-Dreps, Sylvia
  • Zhou, Haibo
  • Zou, Fei
  • Weaver, Mark
  • Koch, Gary
  • Doctor of Public Health
Degree granting institution
  • University of North Carolina at Chapel Hill Graduate School
Graduation year
  • 2017

This work has no parents.