|
Activity Number:
|
76
- Contributed Poster Presentations: Section on Statistics in Epidemiology
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 3, 2020 : 10:00 AM to 2:00 PM
|
|
Sponsor:
|
Section on Statistics in Epidemiology
|
|
Abstract #313779
|
|
|
Title:
|
Doubly Robust Estimators in Randomized Trials with Monotone Dropout
|
|
Author(s):
|
Yuqi Qiu* and Karen Messer
|
|
Companies:
|
and University of California, San Diego
|
|
Keywords:
|
doubly robust estimator;
monotone dropout;
randomized trials;
longitudinal imputation
|
|
Abstract:
|
Monotone dropout due to different reasons in randomized trials may lead to invalid inference when doing analysis for completers only. Robins proposed weighted generalized estimating equations (WGEE) to make consistent inference under missing at random (MAR) but it requires both marginal model and dropout model are correctly specified. Doubly robust (DR) methods have been developed as an alternative for imputing missing data since they give consistent estimators when either the mean imputation model or the dropout model is correct. However, DR methods are rarely used in practice for longitudinal studies and randomized trials due to their complexity. In this paper, existing doubly robust methods in both cross-sectional and longitudinal forms were reviewed, and we proposed a simpler form for longitudinal data. Furthermore, we described the details of making inferences for both outcome and coefficients in longitudinal studies. Then selected DR estimators including the proposed one were compared with other popular imputation methods in several longitudinal simulation studies. We also analyzed data from Alzheimer’s randomized trials with dropout by using DR and other approaches.
|
Authors who are presenting talks have a * after their name.
