425 – Contributed Poster Presentations: Survey Research Methods Section
Competing Imputation Approaches Under Simulated Nonignorable Missingness for Perpetrator Characteristics in the FBI's Supplementary Homicide Reports
G. Lance Couzens
RTI International
Marcus Berzofsky
RTI International
This paper compares two common methods for imputation (Fully Conditional Specification and Weighted Sequential Hot-Deck) under varying levels of simulated non-ignorable missingness in the Federal Bureau of Investigation's (FBI's) Supplementary Homicide Reports (SHRs). The SHR data contain valuable detailed information on victim, perpetrator, and incident characteristics relating to homicides occurring in the U.S., but the utility of these data is limited by the high rates of missingness on perpetrator variables resulting from homicide case insolvency. Particular attention is given to the formulation of the missingness induction mechanism which utilizes information known about victim characteristics in unsolved cases to simulate non-ignorable missingness for known perpetrators. Simulation-based sensitivity results show the methods compare similarly on these data, though neither is able to achieve adequate estimate coverage nor eliminate the directional bias in point estimates given even moderate mechanism strengths.