Details

Mayleen Cortez-Rodriguez, a finalist for a tenure-track faculty position in the Department of Mathematics, will deliver the lecture"Estimating Causal Effects in the Presence of Interference with No Network Knowledge"

Abstract

Do phone call reminders to vote increase voter turnout? Does a vaccine decrease disease rates? The answers to causal questions like these can inform public policy decisions and public health strategies. A broad goal of causal inference, the field that studies causality, is to quantify the causal effect of a treatment (e.g. phone call or vaccine) on an outcome (e.g. voter turnout or disease rate) to answer a causal question of interest. Randomized Control Trials (RCTs) are considered the “gold standard” for estimating causal effects because they satisfy an important independence assumption about the treatment. However, even RCTs can be subject to interference, where the treatment of one individual can affect the outcome of another (e.g., vaccines and herd immunity). Many approaches to addressing interference have arisen in the last decade and most begin by modeling the interference as a network. However, the majority of these approaches require full knowledge of the underlying network, which is a problem in many practical settings where such information is unavailable. Thus, to address this critical gap, Cortez-Rodriguez presents an approach that achieves unbiased estimates of causal effects without requiring fine-grain knowledge of the underlying interference network. Then, she presents a way to incorporate two different types of network information to improve performance. Cortez-Rodriguez concludes with some promising directions for future work.