Statistical Analyses of Victims 1-3 Treatment to Predict Team Versus Solo Serial Killers

Nikki Igo, John Harden, Jason Beaman

Research output: Contribution to journalArticlepeer-review


Study of serial killer teams has remained a blind spot in forensic research despite the persistent popularity of serial homicide as a research topic. Abundant data available through the Radford/FGCU Serial killer allowed for direct comparison of serial killers teams and soloists via statistical analyses of treatment variables of the first three victims in a series. The use of forward binary logistic regression identified the behaviors of stalking, raping, acting quickly, and use of a gun as significant variables. Analysis of these variables using CatPCA resulted in findings that teams are more likely to kill with a gun than soloists and have less tendency to rape, stalk and kill quickly than do their solo counterparts. The aforementioned data analyses culminated in the development of models able to predict whether a homicide was committed by a team or an individual with a range of 79.1-89.1% correct prediction for individuals and 92.1-96.1% for teams, with means of correct prediction for individuals and teams as 83.8% and 92.5%, respectively. Serial Killer Teams are more likely to kill by gun and less likely to kill quickly, stalk, or rape, and the opposite was found for Individual Serial Killers, across all models. These findings suggest great potential for further quantitative analyses of serial killer team data and utility of such information in the examination of homicide cases believed to be committed in series. Furthermore, this novel method of model development could be adapted for prediction models in other areas of forensic study.
Original languageAmerican English
JournalOklahoma State Medical Proceedings
Issue number2
StatePublished - 13 Dec 2021


Dive into the research topics of 'Statistical Analyses of Victims 1-3 Treatment to Predict Team Versus Solo Serial Killers'. Together they form a unique fingerprint.

Cite this