Activity Number:
|
583
- Statistical Applications in Observational Studies
|
Type:
|
Contributed
|
Date/Time:
|
Wednesday, August 2, 2017 : 2:00 PM to 3:50 PM
|
Sponsor:
|
Section on Statistics in Epidemiology
|
Abstract #323669
|
View Presentation
|
Title:
|
On Sensitivity Value of Pair-Matched Observational Studies
|
Author(s):
|
Qingyuan Zhao*
|
Companies:
|
University of Pennsylvania
|
Keywords:
|
effect modification ;
genomics screening ;
signed score test ;
study design ;
U-statistics
|
Abstract:
|
An observational study may be biased by failing to control for unmeasured covariates. A sensitivity analysis asks how susceptible the conclusion of a naive analysis is to possible unmeasured confounders. This paper proposes a new quantity called the "sensitivity value", which is defined as the magnitude of departure from a randomized experiment needed to change the qualitative conclusions. We establish the asymptotic normality of the sensitivity value in pair-matched observational studies. The theoretical results are used to compute the power of a sensitivity analysis and select the design of a study. We explore the potential to use sensitivity values to screen multiple hypotheses using a microarray dataset and find it is robust against unobserved confounding.
|
Authors who are presenting talks have a * after their name.