Omnibus Test in Matched Case-Control Studies

• Activity Number: 325 - Semiparametric Regression in Matched Case-Crossover Studies
• Type: Topic Contributed
• Date/Time: Tuesday, August 1, 2017 : 10:30 AM to 12:20 PM
• Sponsor: ENAR
• Author(s): Byung-Jun Kim* and Inyoung Kim
• Companies: Virginia Polytechnic Inst. & State Univ. Department of Statistics and Virginia Tech
• Keywords: 1-M Matched study ; Measurement error ; Nonparametric regression

Abstract

Matched case-control study is a special case of the stratified case-control study. In design of this study, subjects are stratified by matching on demographic factors such as age and sex. Matched studies have groups of binary-valued outcome indicating case and controls. We use one case and from one to many controls per stratum in the matched design which are called as 1-M matched studies. Our goal of study is to identify the association between covariates with error and clustered binary outcomes. Because the covariates in 1-M matched study are measured with error without specifying distributional assumptions, it is difficult to test whether there are functional association between the clustered outcomes and covariates. In consideration of this non-parametric setting where both the functional form and the distribution of error in covariate are unknown, we propose an omnibus test for the significance of the functional association. Our omnibus test is developed by using estimating equations where likelihood function is not available. The advantage of the omnibus test is demonstrated through analysis of 1-M matched study data set.