Abstract Details
Activity Number:
|
3
|
Type:
|
Invited
|
Date/Time:
|
Sunday, August 3, 2014 : 2:00 PM to 3:50 PM
|
Sponsor:
|
Social Statistics Section
|
Abstract #310774
|
|
Title:
|
Implications of Measurement Error on Covariate Selection for Causal Inference
|
Author(s):
|
Peter M. Steiner*+ and Yongnam Kim
|
Companies:
|
University of Wisconsin-Madison and University of Wisconsin-Madison
|
Keywords:
|
causal inference ;
measurement error ;
propensity score ;
covariate selection
|
Abstract:
|
With observational data, unbiased treatment effects are estimable only if the selection mechanism into treatment conditions is strongly ignorable, that is, if all constructs that are simultaneously related to both treatment selection and the outcome are (i) observed and (ii) measured without error. Using directed acyclic graphs and result from simulation studies, we discuss several issues that directly follow from unreliably measured covariates with respect to causal inference via propensity score methods. First, we show that measurement error attenuates a covariate's potential to remove selection bias. Second, we demonstrate that a set of interrelated covariates can partially compensate for each other's measurement error. Third, balance metrics for specifying the PS model lack power and tend to underestimate the imbalance in unreliably measured covariates. Forth, remaining bias due to measurement error is amplified by (near) instrumental variables. Fifth, investigations of moderation effects (i.e., treatment effect heterogeneity) are generally flawed.
|
Authors who are presenting talks have a * after their name.
Back to the full JSM 2014 program
|
2014 JSM Online Program Home
For information, contact jsm@amstat.org or phone (888) 231-3473.
If you have questions about the Professional Development program, please contact the Education Department.
The views expressed here are those of the individual authors and not necessarily those of the JSM sponsors, their officers, or their staff.
Copyright © American Statistical Association.