Abstract：

Ending the HIV/AIDS epidemic is a goal born from over 30 years of devastation, struggle, and loss, and contains within it hope and promise for those affected. Great progress has been made, including a 38% decline in new infections since 2001, a 58% drop of new infections in children since 2002, and a 35% fall in AIDS-related deaths since 2005. However, this progress does not include many populations and areas. Key populations–such as sex workers, people who inject drugs, transgender people, prisoners, gay men, and other men who have sex with men–remain at much higher risk of HIV infection.

To reach the visionary goal of ending the AIDS epidemic by 2030, policies and interventions require more accurate estimates of the epidemic in both the general population and key at-risk populations, at global and local scale. Size estimation of key population is often a difficult task because they are hard to reach and often hidden, so standard survey and census methods are inadequate. As a result, it is often necessary to leverage multiple data sources, each of which may by itself provide limited information. In addition, expert knowledge about the size of the populations can be useful, even if it is not very precise. We develop a Bayesian hierarchical model for estimating the sizes of local and national key HIV affected populations. The model incorporates multiple commonly used data sources including mapping data, surveys, interventions, capture-recapture data and estimates or guesstimates from organizations, as well as expert knowledge.

It is also important to understand data contributions to the estimation of epidemics. We develop new value of information methods to apply to the problems of outlier detection and influence analysis. The proposed method has a distinct advantage in flexibility and interpretability when compared to existing methods. HIV prevalence in Lesotho and is used to illustrate the use of a value of information approach to influence analysis in the case of a generalized linear mixed model.

Bio：

Dr. Le Bao is an associate professor of Statistics and the associate director of center for advanced data assimilation and predictability techniques at Penn State University. Dr. Bao also serves as the key technical advisor for the UNAIDS Reference Group, which advises on the methods for calculating international AIDS statistics, and as a core project team leader of the Diagnostics Modeling Consortium, which aims to use modeling to guide the effective use of diagnostic technologies in resource-poor settings. Dr. Bao earned his PhD from the Department of Statistics at the University of Washington, Seattle. His research focuses on 1. using statistical models to address global health issues such as estimation of HIV epidemics, health indicators, age and cause specific child mortality; 2. developing fast algorithms for big data; 3. developing methods in categorical data analysis.

http://www.personal.psu.edu/lub14/

欢迎广大老师和同学们参加！