A Kernel Weighting Approach to Improve Population Representativeness for Association Estimation

• Activity Number: 158 - Inference with Non-Probability Sample Through Data Integration
• Type: Topic Contributed
• Date/Time: Monday, July 29, 2019 : 10:30 AM to 12:20 PM
• Sponsor: Survey Research Methods Section
• Author(s): Lingxiao Wang* and Barry Graubard and Hormuzd Katki and Yan Li
• Companies: and National Cancer Institute and US National Cancer Institute and University of Maryland at College Park
• Keywords: Nonprobability Cohort studies; Complex survey sample; Jackknife variance estimation; Kernel smoothing; Propensity score weighting; Taylor series linearization variance

Abstract

In epidemiology, association between risk factors and diseases is important to study for human diseases. There is a vigorous debate about the value of population representative samples for external validity of association estimation. Some literature argues that lack of representativeness may not lead to large bias in estimates of association when the confounders are appropriately controlled. Others advocate the necessity of representative sample. With limited availability of variables, or misspecified analysis models, controlling for confounders may not substantially reduce the bias if the sample is non-representative. This talk will present how non-representative sample may bias the estimates of association in the population and introduce an efficient a kernel weighting (KW) approach to improve the external validity of the association estimation. Jackknife variance estimators are developed to account for all sources of variability of the KW estimators. Monte Carlo simulation studies show that the proposed KW estimators of association overperform the existing inverse propensity-score weighting and subclassification estimators in terms of balance the bias and variance trade-off.