Primary Submission Category: Causal Inference in Networks
Randomized Experiment for Dyadic Data with Interference
Authors: Yilin Li, Lu Deng, Yong Wang, Wang Miao,
Presenting Author: Yilin Li*
Estimating the global average treatment effect (total treatment effect) on a network could be considerably difficult in the presence of unknown network interference. We consider novel setting where the dyadic outcomes are available. Dyadic outcomes are common in many social network sources, such as forwarding a message or sharing a link. We first introduce the setting of network interference with unit-level treatment and dyadic outcomes, which is of particular interest in online experimentation. Then we manifest that the unbiased estimator for the global average treatment effect based on the unit-level outcomes does not exist in general. We provide subsequently unbiased estimators based on dyadic outcomes for randomized experiments. We show the possible variance bounds of our proposed estimators and provide an asymptotic conservative variance estimator. We illustrate the above phenomenon with a variety of numerical experiments. We utilize our method and discuss an application on the WeChat Channel.