The Evaluation of Fibre Transfer Evidence in Forensic Science: A Case Study in Statistical Modelling

Abstract

SUMMARY Frequently, when a crime is committed, fibres are left at the scene. This paper examines the modelling aspects of evaluating the evidential content of such fibres by using a Bayesian approach. Inferences are made via the likelihood ratio, derived from bivariate colour measurements. Modelling the distribution of colour within a particular garment is discussed in detail. In addition, a large database allows an empirical prior distribution to be incorporated, utilizing kernel density estimation. Data from actual casework are analysed.

Bivariate Kurtotic Distributions of Garment Fibre Data

“…Our analysis has shown that the bivariate uniform power distribution is a suitable and numerically tractable model for modelling and comparing the within‐garment variation of the old and new data sets. Our analysis agrees with the findings of Wakefield et al . (1991) in that the within‐garment variation of fibres in the old data set is platykurtic in both axes, but more so in the long axes.…”

“…A preliminary analysis of the within‐garment variation of the old data set by Wakefield et al . (1991) indicated that normal distribution assumptions were not valid in that negative kurtosis (platykurtic shapes) was present.…”

“…Section 2 explores models for within‐garment variation. The models that were considered by Wakefield et al . (1991) and scale mixtures of the normal distribution are briefly reviewed.…”

“…Throughout we take a Bayesian perspective. The applicability of Bayesian statistical methods to forensic identification problems has been established by Lindley (1977) and Evett (1984Evett ( , 1987 and the Bayesian analysis of fibre transfer evidence has been studied by Evett et al (1987) and Wakefield et al (1991). This paper focuses on modelling bivariate colour measurements in the population of garment fibres.…”

Complexity in Forensic Science

“…If the above factors are not taken into account, there is a real danger that the findings produced by the forensic scientist will be misleading. Early in one of the author's careers (Wakefield 1991), she saw many examples of the court over-interpreting the value of her evidence because of the norm at that time to factually report, e.g., three fibers were recovered matching the suspect's jumper or paint matching the window was recovered from the jacket. The reports at that time detailed the analytical work but did not give the court any assistance as to the significance of the findings in the context of the case.…”