Magazine
Global Health, Conflicted Data, and GPS: Analyzing a Gender-Based Violence Intervention in Nairobi, Kenya (303724)
Mike Baiocchi, Stanford University*Rina Friedberg, Stanford University
Clea Sarnquist, Stanford University
Keywords: Protected data, spatial models, global health, missing data
MOTIVATING EXAMPLE: We present novel methods for analyzing highly sensitive information through a worked example: an intervention to lower rates of sexual assault in Nairobi. Self-reported and extremely protected data often contains missing and conflicting reports; we highlight a model to adjudicate problematic responses. This framework can provide insights into how those errors arise during data collection; for us it uncovered a story about locations of violence and community structure. This dynamic was under-recognized in our original study, motivating a mixed-methods pilot - using (1) precision-GPS data fit to spatial models and (2) thick-description interviews - to provide both actionable insights and a way to meaningfully communicate results. AUDIENCE ENGAGEMENT: We interactively build intuition about important features of basic maps and images and link that to early GPS results, highlighting potential for understanding and building communities. GENERAL LEARNING: Our work is widely accessible. Leveraging data to understand human behavior, from individual responses to different ads to how government policies affect society, is a challenge of enormous multidisciplinary interest.