For best experience please turn on javascript and use a modern browser!
You are using a browser that is no longer supported by Microsoft. Please upgrade your browser. The site may not present itself correctly if you continue browsing.

Linking gasoline residues to arson cases: improving methods to assess the strength of evidence

Forensic Expertise Area : Forensic Statistics

In arson cases, often remains of an ignitable liquid are found at the start of the fire. In many cases, at a suspect a jerry can is found with some leftover gasoline. In order to link the gasoline residue in the jerry can to the arson case, a complex chemical analyses proceeds an automated comparison tool. The comparison tool is probabilistic in nature, and the outcome is based on a multivariate analyses for which parameters are trained by several background datasets. The performance of the model is assessed by using an independent validation dataset. 

Assessing the strength of evidence. Two score distributions are depicted. One for which the prosecutor’s hypothesis (gasoline trace at crime scene is from the gasoline in the jerry can) is true and one for which the defense hypothesis (gasoline trace is from unknown different source) is true. For the case, a comparison score is evaluated from chemical analysis and the strength of evidence is assessed by the ratio of the two probability densities at this comparison score.
Assessing the strength of evidence. Two score distributions are depicted. One for which the prosecutor’s hypothesis (gasoline trace at crime scene is from the gasoline in the jerry can) is true and one for which the defense hypothesis (gasoline trace is from unknown different source) is true. For the case, a comparison score is evaluated from chemical analysis and the strength of evidence is assessed by the ratio of the two probability densities at this comparison score.

RESEARCH GOAL

Even though at the moment performance is good, we believe there is room for improvement. Research questions we have are: can the number of parameters be diminished so that the models become simpler and less training data need to be used? How sensitive is the model for choice of validation databases? If results are promising they have the potential to be published in a peer reviewed forensic scientific journal.

REQUIRED/RECOMMENDED EXPERTISE

A background in research methodology, statistics, or anything related to statistics.


 

 Institute/Company :

 Netherlands Forensic Institute
City: Den Haag, The Netherlands
Country :  The Netherlands
Supervisor :  Peter Vergeer

UVA Examiner:

  

UVA Coordinator: 

  
Contact Peter Vergeer
Peter Vergeer
Peter Vergeer