A Comparison of Imputation Methods for Missing Data in Self-Report Likert Ratings

• Activity Number: 285
• Type: Contributed
• Date/Time: Tuesday, August 4, 2009 : 8:30 AM to 10:20 AM
• Sponsor: Section on Health Policy Statistics
• Author(s): Lingqi Tang*+ and Thomas R. Belin and Judy Ho and Bonnie Zima
• Companies: University of California, Los Angeles and University of California, Los Angeles and University of California, Los Angeles and University of California, Los Angeles
• Keywords: missing data ; imputation ; self-report measures

Abstract

In this paper, we compare alternative methods of handling missing data in likert rating scores. The first approach combines two methodologies, utilizing case mean substitution when missing data comprises up to 20% of items, and utilizing sample mean substitution when missing items exceed 20%. A second method is based on a multivariate normal model using PROC MI in SAS software V9.1. These two approaches are contrasted with a pairwise deletion technique. All three methods are applied to a quality of care study for children with Attention Deficit Hyperactivity Disorder (ADHD) in public primary care and managed care Medicaid programs, in which the relationship between parental ADHD knowledge scores, race/ethnicity, and recent service contact are investigated.