|
Activity Number:
|
209
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, July 30, 2007 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Health Policy Statistics
|
| Abstract - #308807 |
|
Title:
|
On the Meta-Analysis of Incomplete Primary Study Data
|
|
Author(s):
|
Todd Bodner*+
|
|
Companies:
|
Portland State University
|
|
Address:
|
5744 SE Preston Ct, Hillsboro, OR, 97123,
|
|
Keywords:
|
Meta-analysis ; Missing data ; Effect sizes
|
|
Abstract:
|
Meta-analysis serves to compare and combine primary study results often stated as effect sizes. However, incomplete data are common in primary research studies. Despite large literatures on meta-analytic and missing data methods, little research has explored the impact of incomplete primary study data on effect sizes and their weights and subsequently on meta-analytic inference under the variety of missing data methods. This poster focuses on the computation of effect sizes and their weights along with their implications for meta-analytic inference under situations with complete and varying degrees of incomplete data handled using either multiple imputation or listwise deletion. Results include the finding that under MCAR these missing data methods can generate large differences in effect size weights depending on the amount of missing data which can impact meta-analytic inference.
|
