Online Program Home
My Program

Abstract Details

Activity Number: 383 - New Developments in Sensitivity Analysis for Unmeasured Confounding
Type: Invited
Date/Time: Tuesday, July 31, 2018 : 2:00 PM to 3:50 PM
Sponsor: Section on Statistics in Epidemiology
Abstract #326804 Presentation
Title: Causal Inference Using a Bayesian Nonparametric Model with Informative Priors on Sensitivity Parameters
Author(s): Jason Roy*
Companies: University of Pennsylvania
Keywords: causal inference ; sensitivity analysis; Bayesian

We propose sensitivity analysis methods for the sequential ignobility (SI) assumption in time-dependent confounding settings. Our general approach is to first model the joint distribution of the observed data using flexible Bayesian nonparametric methods. Static or dynamic causal effects can then be computed using the g-formula. The usual approach is to compute causal effects under the SI assumption We propose some specific forms for departures from sequential ignobility that involve several sensitivity parameters. We propose a revised g-computation step that first involves placing informative priors on these parameters and then computer causal effects that capture uncertainty about the SI assumption. We illustrate the methods using data from a study on prolonged corticosteroid therapy versus anti-tumor necrosis factor alpha directed therapy for treating inflammatory bowel disease.

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

Back to the full JSM 2018 program