Reliability of Relative Standard Errors Computed from NHDS Public Use Data Files

• Activity Number: 191
• Type: Contributed
• Date/Time: Monday, August 2, 2010 : 10:30 AM to 12:20 PM
• Sponsor: Section on Survey Research Methods
• Author(s): Bill Cai*+ and Iris Shimizu
• Companies: National Center for Health Statistics and National Center for Health Statistics
• Keywords: RSE ; NHDS ; SUDAAN ; GVF ; public use data file

Abstract

The National Hospital Discharge Survey (NHDS) is conducted to produce nationally representative estimates of hospital discharges. Micro data files containing data collected in the NHDS are released to the public, but sampling design variables required for computing the NHDS variances are omitted from these files to maintain confidentiality of respondents' identities. To enable data users to compute variances for estimates derived from the public use data files, generalized variance functions (GVFs) are supplied in the public use data file documentation. This paper discusses the quality of the GVF results. It compares using GVFs to derive relative standard errors (RSEs) (aka coefficients of variation or CVs) from the public data use files with applying SUDAAN to in-house data files to obtain RSEs.