Keynote speech 1
Title
:
On testability of causal and missing data models using Verma constraints
Presenter
:
Professor Razieh Nabi, Emory Rollins School of Public Health
Abstract
:
Significant progress has been made in developing identification and estimation
techniques for causal effects in the presence of unmeasured covariates where
modeling assumptions can be described via a directed acyclic graph. For instance,
front-door adjustment (Pearl, 1995) offers an appealing alternative in settings
where standard covariate adjustment is not possible. However, several authors have
cast doubt on whether the assumptions encoded by the model are plausible in
practice. In this talk, we explore the testability of assumptions encoded in the
front-door model (and simple extensions of it) via generalized equality constraints
a.k.a Verma constraints. We propose two goodness-of-fit tests based on this
observation, and evaluate the efficacy of our proposal on real and synthetic data.
We further entertain similar ideas to testability of assumptions encoded in
graphical models of missing data.
Biography
:
Razieh Nabi is a Rollins Assistant Professor in the Department of Biostatistics
and Bioinformatics at Emory Rollins School of Public Health. Her research is
situated at the intersection of machine learning and statistics, focusing on
causal inference and its applications in healthcare and social justice. More
broadly, her work spans problems in causal inference, mediation analysis,
algorithmic fairness, semiparametric inference, graphical models, and missing
data. She has received her PhD (2021) in Computer Science from Johns Hopkins
University.
Keynote speech 2
Title
:
Causal Discovery from Social Networks
Presenter
:
Professor Elena Zheleva, Computer Science, University of Illinois at Chicago
Abstract
:
Social network data breaks a fundamental assumption of existing causal inference techniques, known as the Stable Unit Treatment Value Assumption (SUTVA), which states that the treatment of one unit cannot influence the outcome of other units. In real-world scenarios, it is common for related units to interact with each other which leads to interference (also known as spillover, or peer effects), in which the outcomes of units are interdependent. For example, the opinion of one person can influence the opinion of their friends, and the health status of one individual can impact the health status of others they interact with. In this talk, I will focus on recent work on discovering causal insights from social network data. First, I will present a causal view of a well-known information diffusion model, the linear threshold model, showing how causal inference can help with better threshold and diffusion prediction. Then, I will talk about our work on representation and learning of relational causal models with cycles which is central to causal discovery from observational social network data.
Biography
:
Elena Zheleva is an assistant professor of Computer Science at the University of Illinois at Chicago. Her research focuses on the intersection of causal inference and machine learning for network data, with applications to social media, health, recommender systems, and privacy. Her research is inspired by her industry experience as a data scientist and her public policy experience as a AAAS Science and Technology Policy Fellow. Her research is supported by NSF CAREER, Adobe Research, Anthem and DARPA.