Data Quality Control in Social Surveys Using Genetic Information

摘要

This article introduces a novel way of taking advantage of genetic data in social surveys for the purposes of data quality control. Genetic information could detect and repair data issues such as missing data, reporting errors, differences in measures of the same variable, and flawed data. Using data from two surveys, the College Roommate Study (ROOM) and the National Longitudinal Study of Adolescent Health (Add Health), we show that proportion identical by descent score (a measure of genetic relationships) can identify "misreported" and unreported sibling type and detect misrepresented participants, bio-ancestry score (a measure of ancestral population memberships) can repair and recover missing race and discrepancies among different measures of self-reported race, and sex chromosomal information may help cross-check self-reported sex. This article represents an initial effort to utilize genetic data for the purposes of data quality control. As genetic data become increasingly available, researchers may explore more approaches to improving data quality.