Bayesian modelling for forensic evaluation of handwritten loop characters

This paper presents a multidisciplinary approach to identifying valid and useful patterns in handwriting examination through Bayesian modelling. A sample of loop characters from 13 French writers was analyzed using Fourier analysis to achieve accurate reconstruction. The contour shape of the loop characters is described using Fourier coefficients and surface size. Two modeling approaches are considered: a two-level random effects model [1] and a Bayesian MANOVA model. For both models, two different Bayesian versions with different prior specifications are considered: a conjugate approach and an independent prior approach. Due to the complexity of the latter, Monte Carlo methods are required for estimating marginal likelihoods. The performance of the proposed models was compared using the Bayes factor, evaluating their efficiency in discrimination tasks. The Bayesian MANOVA model, which considere