The evaluation of evidence in the form of multivariate data and in the absence of population data

Lead Research Organisation: University of Edinburgh
Department Name: Sch of Mathematics


Often in the investigation of a crime and in any subsequent trial it is necessary to evaluate so-called trace evidence. Trace evidence is, in a circular definition, evidence which is in the form of traces. Traces may include blood stains, from which a DNA profile may be obtained, and fragments of glass from which measurements of the refractive index and the elemental composition of the glass may be taken. This research is concerned with trace evidence associated with pollen, with minerals, with gun-shot residues and with paint traces.Trace evidence found at a crime scene and on a suspect are compared for two characteristics, similarity and rarity. Generally, similarity and rarity are assessed with reference to some relevant background population. Thus, the DNA profile of a blood stain found at a crime scene is compared with the DNA profile of a suspect. The profiles may match. If they do, it is important to know how rare the profile is in some population.Several problems in evidential assessment are addressed by this research proposal.First, the data are compositional. They measure the proportions of various elements or chemicals that are contained in the evidence (e.g., what proportions of different pollen are contained in the palynological data). Very little research has been done to evaluate evidence of this kind.Second, there is an absence of population data. There are limited control data from the proximity of the crime (e.g. plants to provided pollen for palynology or mineral traces for mineralogy). There will be crime scene data, data from the suspect and control data. Models for the assessment of these need to be developed.These models will be compared using measures of distance. These will depend on the differences found on the data from the various sources.Software will be developed to implement the models investigated in the proposal and the software will be developed in collaboration with practising forensic scientists. There are two stages of a criminal investigation in which the software may be of use. First there is the investigation where the output from the software may help guide the investigating officers as to which lines of enquiry are the most profitable to pursue. Second, there is the evaluation of evidence and the presentation of that evidence in court where the software may help to do this.Finally, care has to be taken that the methods and software developed in this research satisfy the rules of evidence.


10 25 50
publication icon
Mavridis D (2009) Sample size determination for categorical responses. in Journal of forensic sciences

publication icon
Neocleous T (2011) Transformations for compositional data with zeros with an application to forensic evidence evaluation in Chemometrics and Intelligent Laboratory Systems

publication icon
Zadora G (2010) A two-level model for evidence evaluation in the presence of zeros. in Journal of forensic sciences

publication icon
Zadora G (2009) Likelihood ratio model for classification of forensic evidence. in Analytica chimica acta