Activity Number:
|
611
|
Type:
|
Contributed
|
Date/Time:
|
Wednesday, August 3, 2016 : 2:00 PM to 3:50 PM
|
Sponsor:
|
Biopharmaceutical Section
|
Abstract #319562
|
|
Title:
|
On the Use of the Treatment Effect in the Imputation Model for Multiple Imputation Analyses of Missing Data
|
Author(s):
|
Robert Small*
|
Companies:
|
Sanofi Pasteur
|
Keywords:
|
Multiple Imputation ;
Missing data ;
Imputation Model
|
Abstract:
|
Various authors have debated the use of the treatment effect in an imputation model for Multiple Imputation. Some have followed the advice of putting as much into the imputation model as available since we may not know the relationship between observed covariates and the missing data. But the treatment effect is a unique independent variable and the objective of the original experiment. It is central to the randomization that defines the experiment and an unfailing justification for the analyses. In this paper we define a simple model with a MCR missing pattern. The ML estimates and tests are available. We compare ML approach, the MI approach and a randomization approach analytically and with simulations.
|
Authors who are presenting talks have a * after their name.
Back to the full JSM 2016 program
|