Statistics and Data Science Seminar - Yuhua Zhang
— 5:00pm
Location:
In Person
-
Steinberg Auditorium, Baker Hall A53
Speaker:
YUHUA ZHANG
,
Postdoctoral Fellow, Department of Biostatistics, Harvard University
https://yuhuazhang1995.github.io/
Networks are everywhere, from patient transfers across hospitals to interactions on social platforms. Yet such data is not fully explored. If we can learn how entities connect and influence one another, we can identify population disparities in complex systems and design interventions that improve social and public health outcomes. In this talk, I will cover two complementary themes, modeling of network-structured data and causal inference in networks, and show how network structure can be translated into actionable scientific and policy insight.
In the first part, I will discuss community detection in networks. While many methods aim to recover community structure, fewer account for the fact that modern networks often arise from interaction processes, where edges, not nodes, are the basic statistical units. I will present block edge exchangeable models (BEEM) for interaction networks with latent node-level community structure, and show how this model enables inference for sparse network data.
In the second part, I will discuss causal inference with unknown interference networks. Interference, the phenomenon that a unit’s outcome depends on other units’ treatments, creates major challenges for identifying and estimating causal effects. Most existing approaches assume the interference network is known, which is often unrealistic in practice because such a network is typically latent or only partially observed. To address this, I develop a framework for identifying and estimating heterogeneous, group-level interference effects without requiring a known interference network. I will illustrate these ideas with applications to healthcare systems, social networks, and more.
—
Yuhua Zhang is a postdoctoral fellow working with Dr. Jukka-Pekka Onnela in the Department of Biostatistics at Harvard University. Prior to this, she earned her Ph.D. from the Department of Biostatistics at the University of Michigan, supervised by Dr. Walter Dempsey and Dr. Sebastian Zoellner. Her research interests include network analysis, causal inference in networks and their applications in social and health science.
For More Information:
jpaschke@andrew.cmu.edu