Evaluating the Effect of Weighting Adjustments on HIV Prevalence Estimates in Three Population-Based HIV Impact Assessments Surveys

• Activity Number: 301 - Weighting Adjustments
• Type: Contributed
• Date/Time: Tuesday, August 1, 2017 : 8:30 AM to 10:20 AM
• Sponsor: Survey Research Methods Section
• Author(s): Tien-Huan Lin* and Natalia Weil and Ismael Flores Cervantes and Suzue Saito
• Companies: Westat and Westat and Westat and ICAP at Columbia University
• Keywords: Survey weights ; nonresponse ; nonresponse weighting adjustment ; LASSO regression ; variable selection ; CHAID

Abstract

Population-based HIV Impact Assessment (PHIA) surveys are being conducted in 12 sub-Saharan African countries to measure HIV incidence, prevalence, and other key impact indicators by ICAP at Columbia University in collaboration with ministries of health and the Centers of Disease Control and Prevention and other partners. We use data from the first three PHIA surveys -in Zimbabwe, Malawi, and Zambia- to study the impact of survey weights on estimates of the prevalence of HIV. In developing the survey weights, decisions made about the variables to use for nonresponse adjustments, and about the formation of nonresponse and poststratification cells, may affect survey estimates and their standard errors. The nonresponse adjustments in PHIA surveys include adjustments for household and person nonresponse, and for failure to obtain analyzable blood samples. The paper describes the use of LASSO and CHAID for variable selection in making the nonresponse adjustments. It then examines the effects on selected survey estimates of the use of different sets of variables employed in the nonresponse adjustments and of the use of alternative poststratification adjustments.