The Impact of Missing Data Pattern on Multilevel Mixed-Effects Linear Models

• Activity Number: 336
• Type: Contributed
• Date/Time: Wednesday, August 6, 2003 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Yann-Yann Shieh*+ and Rachel T. Fouladi
• Companies: and University of Texas M.D. Anderson Cancer Center
• Keywords: missing data ; imputation ; multilevel model

Abstract

The problem of missing data arises frequently in practice in applied Research settings. Roth's (1994) review of 75 studies published in the Journal of Applied Psychology and Personnel Psychology indicated that listwise deletion and pairwise deletion were the most common methods of treating missing data. Little and Rubin (1987) noted that "neither method...is generally satisfactory". The biggest drawback associated with ad hoc missing data methods is the strict assumption that the observed data are missing completely at random (MCAR). As pointed out by Schafer and Olsen (1998), methods for analyzing incomplete data have undergone substantial developments in the past 20 years. A key point that is clear from the missing data literature is that one has to choose a computational method or combination of methods based on the nature of the problem, the computational resources, the accuracy requirement, and the degree of difficulty of any required theoretical derivations. In the current paper, the use of multilevel linear modeling under a variety of missing data patterns is contrasted with modeling using missing value imputation methods on model parameter and standard errors.