NotesFAQContact Us
Collection
Advanced
Search Tips
ERIC Number: ED547049
Record Type: Non-Journal
Publication Date: 2012
Pages: 165
Abstractor: As Provided
Reference Count: N/A
ISBN: 978-1-2675-2433-1
ISSN: N/A
The Impact of Five Missing Data Treatments on a Cross-Classified Random Effects Model
Hoelzle, Braden R.
ProQuest LLC, Ph.D. Dissertation, Southern Methodist University
The present study compared the performance of five missing data treatment methods within a Cross-Classified Random Effects Model environment under various levels and patterns of missing data given a specified sample size. Prior research has shown the varying effect of missing data treatment options within the context of numerous statistical models. However, only one study has explored the impact of a missing data treatment within a Cross-Classified Random Effects Model environment and this study only examined the effect of one missing data treatment method. The current study fills a gap in the literature by exploring the comparative effect of five missing data treatment methods (list-wise deletion, mean substitution, group mean substitution, multiple imputation using Bayesian regression, and multiple imputation using the Multilevel Gibbs Sampler method) on the magnitude of the parameter and standard error bias produced by the Cross-Classified Random Effects Model when varying levels and patterns of missing data are simulated for a level-one variable. The results of this study serve as a reference for researchers who are faced with the challenge of determining how to treat missing data when modeling cross-classified data with the Cross-Classified Random Effects Model. A 10,000 iteration Monte Carlo Simulation was conducted on 75 conditions derived from fully crossing the five missing data treatments, five levels of missingness (5%, 10%, 15%, 30%, and 45%), and the three patterns of missingness (missing completely at random, missing at random, and missing not at random). The current study found that list-wise deletion did the best job of mitigating the bias produced by missing data. Multiple imputation using Bayesian regression, mean substitution, and group mean substitution performed adequately and similarly to each other and multiple imputation using the Multilevel Gibbs Sampler method performed poorly. [The dissertation citations contained here are published with the permission of ProQuest LLC. Further reproduction is prohibited without permission. Copies of dissertations may be obtained by Telephone (800) 1-800-521-0600. Web page: http://www.proquest.com/en-US/products/dissertations/individuals.shtml.]
ProQuest LLC. 789 East Eisenhower Parkway, P.O. Box 1346, Ann Arbor, MI 48106. Tel: 800-521-0600; Web site: http://www.proquest.com/en-US/products/dissertations/individuals.shtml
Publication Type: Dissertations/Theses - Doctoral Dissertations
Education Level: N/A
Audience: N/A
Language: English
Sponsor: N/A
Authoring Institution: N/A