Development of new methodologies for dating in the forensic field, combining analytical techniques with multivariate regression treatments

Abstract

Whenever a crime is committed, there is always a unity of time, place and action that forensic experts will aim to demonstrate throughout the investigation. Determining the succession, simultaneity, frequency or duration of criminal activities as well as the age of objects, persons and traces is therefore one of the most important goals of forensics in reconstructing the crime scene or in finding and understanding the connections between the evidence and the suspects involved therein. However, time has been largely unexplored due to the complexity of the overall challenge at hand, not yet successfully overcome by single cutting-edge techniques. The coupling of such techniques with chemometrics, more specifically with multivariate regression methods, could turn this situation upside down thanks to the development of age quantitation methodologies based on the modelling of the modifications experienced by the evidence in its properties with respect to time. The potential applicability of these chemometric tools, however, remains poorly understood and underexploited due to their recent introduction into forensic dating research and the statistical background required for their optimal application. That is why this thesis focuses on highlighting the usefulness of multivariate regression methods in several forensicfields, such as questioned documents, art forgery and medico-legal death investigation, through the development and validation of dating methodologies in which non-destructive and micro-destructive techniques are applied together with the (orthogonal) partial least squares regression ((O)PLSR) method.