The DHS Program

Design-based Small Area Estimation: A Nearest Neighbor Method for HIV Prevalence Estimation (English) 

Document Type

Working Papers

Publication Topic(s)

HIV/AIDS

Country(s)

Rwanda

Language

English

Recommended Citation

Ren, Ruilin. 2022. Design-based Small Area Estimation: A Nearest Neighbor Method for HIV Prevalence Estimation. DHS Working Papers No. 189. Rockville, Maryland, USA: ICF

Download Citation

RIS format / Text format / Endnote format

Publication Date

September 2022

Publication ID

WP189

Abstract:

In 2021, Ren proposed a design-based methodology on small area estimation, the nearest neighbor method, based on survey data collected in a single target survey or in a series of similar surveys conducted in recent years from the same population. In household- based surveys, sampling units geographically close to the small area or sampling units from a social and economic environment similar to the small area may share similar characteristics. The concept involves pooling these sampling units together with the sampling units selected from the small area to form a “nearest neighborhood” with adequate sample size, and then to treat this as a survey domain that represents the small area. A survey domain created in this way is easier to analyze based on the survey design, variance estimation, and confidence intervals. This study is an application of the methods proposed by Ren (2021), in which Ren proposed five nearest neighbor methods: the time-space method, cluster center method, district center method, nearest neighbor with composite distance measure, and the hybrid nearest neighbor. In the 2021 study, Ren applied the proposed methods to produce district level total fertility rates and childhood mortality rates by using the Rwanda 2010 and 2014–2015 DHS survey data. In this study, we use the same survey data and apply the proposed methods to produce district level HIV prevalence estimates for women and men age 15–49. These data were not published in the 2014–2015 survey report because of the insufficient sample size at the district level. The methods are also applicable to HIV prevalence by single sex with an increased neighborhood size.