Oklahoma CODES Crash Outcome Data Evaluation System College of Continuing Education University of Oklahoma

Multiple Imputation

Multiple imputation replaces each missing value with a set of plausible values. These multiply imputed data sets are then analyzed using standard statistical tests. The results from the statistical tests on the multple imputed data sets are then combined. This results in statistically valid inferences that reflect the uncertainty due to missing values. Multiple imputation is also used by CODES to impute mising linkages.

Resources:

• Brief summary of multiple imputation from the SAS website
• The multiple imputation Frequently Asked Question (FAQ) page
• Multiple Imputation for Missing Data: Concepts and New Development (pdf) by Yang C. Yuan, SAS Institute Inc., Rockville, MD
• SAS MI Proceure documentation (pdf)
• SAS MIANALYZE procedure documentation (pdf)
• Imputation of Missing Blood Alcohol Concentration (BAC) Data in the Fatality Analysis Reporting System (FARS) (pdf of NHTSA powerpoint presentation)
• Transitioning to Multiple Imputation – A New Method to Impute Missing Blood Alcohol Concentration (BAC) values in FARS (pdf of NHTSA Technical Report)
• CODES Technical Forum Notes