Evaluation of Two-Step Nonresponse Adjustment in the Mental Health Surveillance Study

• Activity Number: 671
• Type: Contributed
• Date/Time: Thursday, August 2, 2012 : 10:30 AM to 12:20 PM
• Sponsor: Section on Survey Research Methods
• Author(s): Emilia Peytcheva*+ and Phillip Kott and Dan Liao and Kevin Wang and Art Hughes
• Companies: RTI International and RTI International and RTI International and RTI International and SAMHSA
• Keywords: nonresponse ; two-step adjustment

Abstract

Weighting for nonresponse is often done using a single model in which refusals and noncontacts are grouped together. Nonresponse to a survey can occur at multiple stages of contact. When an existing survey serves as a sampling frame there is a rich set of covariates that can be used in adjustments based on a sequence of propensity models. The Mental Health Surveillance Study (MHSS) uses clinical interviews administered to a subsample of respondents to the National Survey on Drug Use and Health (NSDUH) within 4 weeks of completing the NSDUH interview. Two distinct sets of MHSS nonrespondents are observed - those who immediately refuse to participate at the end of the NSDUH interview (initial refusals), and those who initially agree to participate, but cannot be contacted or refuse when contacted for the clinical interview. A comparison of nonresponse adjusted estimates using a single model and a two-step approach, where initial refusals are modeled separately from noncontacts and later refusals, reveals potential advantages of the two-step method. The two-step approach has the potential to yield less biased estimates without appreciably increasing standard errors.