This work presents a novel, multidisciplinary approach that aims to the identification of valid and useful patterns in handwriting examination via Bayesian modelling. Starting from a sample of characters selected among 13 French native speakers, an accurate loop reconstruction can be achieved by means of Fourier analysis. The contour shape of handwritten characters can be described by means of the first four couples of Fourier coefficient and by the surface. Two modelling approaches are considered for such handwritten features: (a) a two-level random effects model proposed by Bozza et al. (2008) and (b) 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. The latter version of these two models is of primary interest because it can incorporate the between writers variability, which is a peculiar distinguishing element between writers, and which is not modelled otherwise. On the other hand, however, this approach does not allow the marginal likelihood to be obtained in closed form, and Monte Carlo methods must be implemented for its estimation.

The Bayes factor is calculated to compare the performance of the proposed models and to evaluate their efficiency for discrimating purposes. Bayesian MANOVA showed an overall better performance, both in terms of stronger evidence and discriminatory capacity. Finally, a sensitivity analysis to the elicitation of the prior distribution modelling the within writers’ variability is performed.