Missing Data in the SHR: Evaluating Multiple Imputation Approaches

• In Event: Modeling and Analytical Issues

Abstract

Criminologists frequently confront missing data in their analyses, and for some time have routinely applied imputation methods as a remedy. An earlier investigation (Wadsworth and Roberts 2008) considered missing data in the Supplementary Homicide Reports (SHR). That work asked how successful existing single imputation approaches were in the SHR context, and employed a unique dataset permitting comparison of data reported as missing (and then imputed) in the SHR to complete data available in police reports for the same homicide incidents. In the time since that research was conducted, multiple imputation approaches have become increasingly popular, and now see wide application in criminology. The present study therefore considers multiple imputation approaches to missing SHR data, again taking advantage of unique comparison data. We discuss technical comparisons as well as assessing the impact of multiple imputation strategies on substantive research conclusions.

Authors

• John M. Roberts, Jr., University of Wisconsin - Milwaukee

• Aki Roberts, University of Wisconsin - Milwaukee

• Tim Wadsworth, University of Colorado, Boulder