Abstract:
|
Public use files (PUFs) for the National Survey on Drug Use and Health (NSDUH) are created through a series of processes, including subsampling, substitution, suppression, top and bottom coding, and collapsing variables. It is important to gauge the impact of this disclosure avoidance treatment on the quality of the data because NSDUH's PUFs are used by researchers to make policy and public health decisions. In this paper, key outcome measures are compared based on NSDUH-restricted files and NSDUH's PUFs. About 200 tables of substance use and mental health estimates from the 2002 to 2013 NSDUHs and their respective standard errors (SEs) were compared. Summary statistics were produced of the ratios of estimates and SEs from the PUF data and the published tables. In addition, global utility measures, such as Hellinger's distance and relative root mean square error, were compared across years for some outcome and domain combinations. This paper demonstrates the overall quality of NSDUH's PUFs and mentions any limitations that were identified, so researchers and other policy makers can continue to derive sound policy-relevant and empirically based conclusions from the PUFs.
|
ASA Meetings Department
732 North Washington Street, Alexandria, VA 22314
(703) 684-1221 • meetings@amstat.org
Copyright © American Statistical Association.