Validation, replication, and sensitivity testing of Heckman-type selection models to adjust estimates of HIV prevalence

Samuel J. Clark*, Brian Houle

*Corresponding author for this work

    Research output: Contribution to journalArticlepeer-review

    32 Citations (Scopus)

    Abstract

    A recent study using Heckman-type selection models to adjust for non-response in the Zambia 2007 Demographic and Health Survey (DHS) found a large correction in HIV prevalence for males. We aim to validate this finding, replicate the adjustment approach in other DHSs, apply the adjustment approach in an external empirical context, and assess the robustness of the technique to different adjustment approaches. We used 6 DHSs, and an HIV prevalence study from rural South Africa to validate and replicate the adjustment approach. We also developed an alternative, systematic model of selection processes and applied it to all surveys. We decomposed corrections from both approaches into rate change and age-structure change components. We are able to reproduce the adjustment approach for the 2007 Zambia DHS and derive results comparable with the original findings. We are able to replicate applying the approach in several other DHSs. The approach also yields reasonable adjustments for a survey in rural South Africa. The technique is relatively robust to how the adjustment approach is specified. The Heckman selection model is a useful tool for assessing the possibility and extent of selection bias in HIV prevalence estimates from sample surveys.

    Original languageEnglish
    Article numbere112563
    JournalPLoS ONE
    Volume9
    Issue number11
    DOIs
    Publication statusPublished - 17 Nov 2014

    Fingerprint

    Dive into the research topics of 'Validation, replication, and sensitivity testing of Heckman-type selection models to adjust estimates of HIV prevalence'. Together they form a unique fingerprint.

    Cite this