Instrumental Variables Methods Reconcile Disparate Intention-to-Treat Effects Across Pragmatic Trials
Principal Investigator
Name
Joshua Angrist
Degrees
PhD
Institution
Massachusetts Institute of Technology
Position Title
Ford Professor of Economics
Email
About this CDAS Project
Study
PLCO
(Learn more about this study)
Project ID
PLCO-1162
Initial CDAS Request Approval
Feb 8, 2023
Title
Instrumental Variables Methods Reconcile Disparate Intention-to-Treat Effects Across Pragmatic Trials
Summary
This project will examine the causal effect of sigmoidoscopy on colorectal cancer incidence and mortality by leveraging the randomization embedded in the PLCO trial. Previous literature has reported intention-to-treat effects in multiple sigmoidoscopy randomized control trials and there has been considerable disparity in the reported effects across different trials. In this study, we will use instrumental variable methods to estimate the effect of undergoing sigmoidoscopy screening. In particular, not all participants who were randomly chosen for screening underwent sigmoidoscopy screening and many offers were declined. Moreover, the adherence rates vary across different trials. We will use the offer to undergo screening to instrument actual sigmoidoscopy screening. We also plan to compare the sample characteristics with compiler characteristics and treated characteristics to understand heterogeneity in compliance.
Aims
- The goal of the project is first to understand the causal effect of sigmoidoscopy on colorectal cancer incidence.
- The project aims to show that an instrumental variable method reconciles disparate intention-to-treat effects across different sigmoidoscopy RCTs with partial compliance.
- We also aim to analyze compiler heterogeneity in these trials.
Collaborators
Peter Hull, Brown University