A Kernel Weighting Approach to Improve Population Representativeness of Epidemiological Cohort in the Analysis

• Activity Number: 555 - Using Surveys to Improve the Representativeness of Nonprobability Samples in Epidemiologic Studies
• Type: Invited
• Date/Time: Wednesday, August 1, 2018 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Lingxiao Wang* and Barry Ira Graubard and Hormuzd A. Katki and Yan Li
• Companies: The Joint Program in Survey Methodology, University of Maryland, College Park and National Cancer Institute and Biostatistics Branch, Division of Cancer Epidemiology & Genetics, National Cancer Institute and University of Maryland at College Park
• Keywords: Non-probability epidemiological cohort; weighting; propensity score; kernel smoothing; complex survey sample; Jackknife variance estimation

Abstract

Epidemiological cohorts are often collected from convenient samples, suffering from selection bias and coverage issue. Accordingly, the analyses are lack of external validity, leading to biased point estimation and invalid variance estimation. In this paper, We developed an efficient kernel weighting (KW) approach that treats a survey sample as a reference to create pseudo weights for the cohort by using propensity scores and kernel smoothing techniques. The proposed jackknife (JK) variance estimator, in addition to accounting for the intracluster correlation induced by the homogeneity of the participants from the same study center, considers the variability due to estimating propensity scores. Monte Carlo simulation studies show that the proposed kernel-weighed estimators reduce the bias and increase the efficiency of the estimated disease prevalence comparing with the existing inverse propensity score weighting (IPSW) method. The proposed JK variance estimators are accurate for variance of IPSW and KW estimates. The developed approach is demonstrated using National Health Interview Survey and NIH-AARP cohort to estimate nine-year all-cause mortality.