Causal Inference Estimation When Treatment Is Continuous Using the Rubin Potential Outcomes Framework

• Activity Number: 230
• Type: Contributed
• Date/Time: Monday, August 4, 2014 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Douglas Galagate*+
• Keywords: causal inference ; missing data ; continuous treatment ; randomized trial ; propensity score

Abstract

Implementing randomized experiments is not always possible in order to estimate causal effects, but an alternative tactic is to use observational data. There are many approaches to estimating causal effects under the binary treatment setting but fewer known methods focus on continuous treatments or multiple treatments. It is possible to imagine a randomized trial with a continuous treatment with the goal of estimating a causal dose-response curve. Estimating the causal effect can be thought of as a missing data problem and we use the Rubin potential outcomes framework as a guide. We compare the current methods for the continuous treatment setting using simulated data and suggest possible improvements.