Online Program Home
My Program

Abstract Details

Activity Number: 585 - Exploiting Latent Structure for Network Inference
Type: Invited
Date/Time: Thursday, August 1, 2019 : 8:30 AM to 10:20 AM
Sponsor: Section on Statistical Computing
Abstract #300602
Title: Leveraging Exchangeability Assumptions to Make Inference in Regression with Network Outcomes
Author(s): Bailey Fosdick*
Companies: Colorado State University
Keywords: relational data; network regression; exchangeability; latent variable models

A common approach to capturing dependencies among network relations in a regression model is to carefully specify a complex latent variable model. Model selection for the latent variable formulation is difficult, yet critical to ensuring accurate inference. In this talk, we move away from focusing on specific latent variable models to considering what implicit assumptions are being made by specifying such a model. Specifically, we consider the assumption of joint exchangeability and explore the implications of this on standard error estimates for the regression coefficients. In the event, joint exchangeability is too restrictive, we introduce relaxations of this assumption where only subsets of the actors are exchangeable.

Authors who are presenting talks have a * after their name.

Back to the full JSM 2019 program