Fig. 1. Example of effect of accounting for nonsampling error in ANC-SS prevalence for Botswana Urban and Zimbabwe Manicaland regions.

Left panels illustrate estimates for ${\sigma }_{infl}^2$ by different approaches. Center panels show the effect on estimates of adult HIV prevalence when ${\sigma }_{infl}^2$ is estimated in full Bayesian inference (pink) versus not included ( ${\sigma }_{infl}^2=0$; green). Right panels illustrate the estimated trend in HIV incidence. Shaded regions represent 95% credible intervals. Results for other approaches and regions are provided in Supplementary Appendix (http://links.lww.com/QAD/B45). ANC-SS, antenatal clinic sentinel surveillance.