Social networks

My work develops statistical models that leverage social network structure to understand the nuances of human behavior. I'm particularly interested in developing methods for partially observed or sampled network data. These data take many forms, ranging from questions asked to respondents on traditional surveys to egocentric subgraphs sampled from social media. Statistical challenges involve modeling the dependence structure introduced through the network and accounting for complex sampling designs or missingness patterns (in both network structure and actor attributes).

Selected publications (for a full list see here):

Estimating vital rates using sparsely sampled data

In most of the developing world, there is massive uncertainty about even the most basic population indicators. These indicators (such as birth rates or disease prevalence) are critical for evaluation and planning of social and public health programs. My work develops statistical models that leverage limited available information to produce estimates that adequately reflect uncertainty arising from various stages of data collection.

Selected publications (for a full list see here):

Predictive models with unsolicited data

Unsolicited, observational data (such as electronic healthcare records) provide exciting opportunities for data-driven decision making. These data arise with no survey or other intervention from researchers and, thus, are subject to substantial selection and reporting issues. My work develops predictive models that leverage the rich information in these data, especially in the context of patient-level predictive models or personalized medicine.

Selected publications (for a full list see here):