EE Systems Seminar: 5th August at 12:00 noon = Causal inference with misspecified network interference | Prof. Daniel Nevo (TAU)
(The talk will be given in English)
Speaker: Prof. Daniel Nevo
Department of Statistics and Operations Research, Tel Aviv University
|
011 hall, Electrical Engineering-Kitot Building |
Monday, August 5th, 2024
12:00 - 13:00
|
|
Causal inference with misspecified network interference
Abstract
The typical approach towards drawing causal conclusions from observed data starts by defining a causal estimand, for example in terms of potential outcomes or the so-called do operator, and continues by providing conditions for identification of this estimand from the data, followed by statistical estimation and inference. One of the main assumptions is the no-interference assumption, meaning that the treatment assigned to one unit does not affect other units in the sample. However, in many domains such as in the social sciences and infectious disease epidemiology, this assumption is implausible in practice due to social interactions.
As an alternative to the no-interference assumption, an interference structure is often represented using a network. Ubiquitously, the network structure is assumed to be known and correctly specified. Nevertheless, correctly encoding the interference structure in a network can be challenging. For example, people may misreport their social connections, or report connections irrelevant to the specific combination of treatment and outcome.
Building on the exposure mapping framework, we derive the bias arising from estimating causal effects under a misspecified interference structure. To address this problem, we propose a novel estimator that uses multiple networks simultaneously and is unbiased if one of the networks correctly represents the interference structure, thus providing robustness to the network specification. Additionally, we propose a sensitivity analysis that quantifies the impact of a postulated misspecification mechanism on the causal estimates. Through simulation studies, we illustrate the bias from assuming an incorrect network and show the bias-variance tradeoff of our proposed network-misspecification-robust estimator. We further demonstrate the utility of our methods in two real examples.
Joint work with Bar Weinstein
Short Bio
Daniel Nevo is an Associate Professor at the Department of Statistics and Operations Research in Tel Aviv University. He received PhD in Statistics in 2016 from The Hebrew University, working on high-dimensional regression problems and biostatistical methods. Then he was a postdoc at Harvard Departments of Biostatistics and Epidemiology, before joining TAU in 2018. Daniel's research focuses on developing and implementing statistical methods in widespread domains, with emphasis on theory and methods for causal inference.
השתתפות בסמינר תיתן קרדיט שמיעה = עפ"י רישום שם מלא + מספר ת.ז. בטופס הנוכחות שיועבר באולם במהלך הסמינר
