|
Activity Number:
|
151
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, July 30, 2007 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Statistics in Epidemiology
|
| Abstract - #309614 |
|
Title:
|
Estimation of Causal Effects in Studies with Outcome-Dependent, Two-Phase Sampling
|
|
Author(s):
|
Weiwei Wang*+ and Daniel Scharfstein and Zhiqiang Tan and Ellen MacKenzie
|
|
Companies:
|
Johns Hopkins Bloomberg School of Public Health and Johns Hopkins Bloomberg School of Public Health and Johns Hopkins Bloomberg School of Public Health and Johns Hopkins Bloomberg School of Public Health
|
|
Address:
|
615 N Wolfe St, Baltimore, MD, 21205,
|
|
Keywords:
|
outcome dependent ; two-phase ; causal inference ; biased sampling ; semiparametric ; doubly robust
|
|
Abstract:
|
We consider studies in which it is inexpensive to measure an outcome Y, a nonrandomized binary treatment T and a subset of confounding factors V, but expensive to measure additional confounding factors W. In such studies, outcome-dependent two-phase sampling can significantly reduce the cost. We propose two estimators of the causal effect of treatment: doubly robust and locally efficient, and compare them to the simple inverse weighted estimator. We illustrate our methods with data from the National Study on the Costs and Outcomes of Trauma and demonstrate finite sample performance of the estimators in a simulation study. We argue that the doubly robust estimator provides the best tradeoff in terms of robustness, efficiency and ease of implementation.
|
