Online Program Home
My Program

Abstract Details

Activity Number: 309 - Advances in Causal Inference
Type: Contributed
Date/Time: Tuesday, July 30, 2019 : 8:30 AM to 10:20 AM
Sponsor: Section on Statistics in Epidemiology
Abstract #302960
Title: Can We Attribute Suicides to an App? Nonparametric Estimation the Probability of Causation
Author(s): Maria Cuellar* and Walter Dempsey
Companies: Carnegie Mellon University and Harvard University
Keywords: causation; probability of causation; attribution; suicide; influence functions; nonparametrics

Researchers often need to determine whether an individual's outcome can be attributed to a specific exposure with some probability. To do this, we suggest estimating the probability of causation (PC). We estimate PC using data from a RCT that uses an application connected to social media that provides mental health support to individuals who have grave depression, as a treatment. The concern posed by a researcher who created the app is: By reminding the user repeatedly not to commit suicide, the app might actually be inadvertently causing the individuals to want to commit suicide, or have some other similar adverse effects. We evaluate the probability that individuals' suicide attempts (and other effects) were caused by the RCT's treatment, not by something else. We use an estimator for PC from our recent work (Cuellar and Kennedy 2018), which is a nonparametric influence-function projection-based (IFB) estimator. This estimator allows for simple interpretation and valid inference by making only weak structural assumptions. We also provide the estimates for the ATE, and we provide parametric and nonparametric IFB estimates for comparison.

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

Back to the full JSM 2019 program