|
Activity Number:
|
216
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 3, 2009 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Survey Research Methods
|
| Abstract - #304189 |
|
Title:
|
Performance of Sequential Imputation Method in Multilevel Applications
|
|
Author(s):
|
Enxu Zhao*+ and Recai M. Yucel
|
|
Companies:
|
New York State Department of Health and State University of New York at Albany
|
|
Address:
|
800 North Pearl Street, Albany, NY, 12204,
|
|
Keywords:
|
missing-data ; multiple imputation ; sequential imputation ; hierarchical models ; random-effects ; clustered data
|
|
Abstract:
|
In missing-data problems complicated not only by the underlying data-generating mechanism (e.g. multi-stage surveys) but also skip patterns or censored items, an imputation strategy proceeding sequentially through the variables has been gaining a great momentum among practitioners. While this approach offers unmatched attractions, it does not necessarily follow a conventional joint modeling approach. Our work assesses the performance and compatibility of sequential approach to a joint modeling approach in multilevel settings. This sequential approach uses a set of hierarchical regression models each of which follows the appropriate format for the underlying variable subject to missingness. A comprehensive simulation study will also be presented, where the performance of the sequential approach is reasonably well and even better than the more convention methods in some scenarios.
|
