Activity Number:
|
531
- SPEED: Statistics in Epidemiology and Genomics and Genetics
|
Type:
|
Contributed
|
Date/Time:
|
Wednesday, August 2, 2017 : 11:35 AM to 12:20 PM
|
Sponsor:
|
Section on Statistics in Epidemiology
|
Abstract #325402
|
|
Title:
|
Estimating Causal Effects from Using Augmented Inverse Probability Weighted Estimators When Noncompliance Is Measured with Error
|
Author(s):
|
David Vock* and Jeffrey Boatman and Joseph Koopmeiners
|
Companies:
|
University of Minnesota and and University of Minnesota
|
Keywords:
|
noncompliance ;
augmented inverse probability weighting ;
clinical trials ;
regulatory tobacco research ;
causal inference
|
Abstract:
|
Noncompliance to randomized treatment is a challenge when interpreting data from randomized trials. Many methods to estimate the causal effect have been proposed, but they typically assume that participants' compliance statuses are reported without error. This is an untenable assumption when noncompliance is self-reported. Biomarkers of exposure may provide more reliable indicators of compliance but cannot perfectly discriminate between compliers and non-compliers. However, by modeling the distribution of the biomarker and response as a mixture distribution and writing the probability of compliance as a function of the mixture components, we show the probability of compliance can be directly estimated from the data when compliance status is unknown. This allows us to derive a class of consistent and asymptotically normal augmented inverse probability of compliance weighted estimators to estimate the causal effect. We demonstrate via simulation that the proposed estimator achieves smaller bias and greater efficiency than ad hoc approaches to estimating the causal effect with unknown compliance status. We apply our method to data from a trial of nicotine content in cigarettes.
|
Authors who are presenting talks have a * after their name.