Workshop

The weight of evidence: Likelihood ratios in forensic statistics

Prof. Giulia Cereda, Ph.D.
Professor of Statistics
University of Florence, Florence, Italy
Homepage

Abstract [Wednesday, September 23, 9.00–13.00]
How should we quantify the strength of forensic evidence? While statistical methods are widely used in forensic science, their interpretation is not always straightforward, and different modelling choices can lead to substantially different conclusions.
This workshop presents likelihood ratios (LR) as a coherent framework for evaluating the weight of evidence, and discusses their role within forensic statistics. The focus is on practical aspects: how likelihood ratios are constructed, how they should be interpreted, and how they behave under different modelling assumptions.
Using examples primarily drawn from forensic DNA analysis, we illustrate how key choices—such as the definition of the relevant population, the treatment of uncertainty, and the handling of dependence—directly affect the evaluation of evidence. These examples highlight an important point: the strength of the evidence is not an intrinsic property of the data, but depends on the underlying statistical model.
We also discuss common pitfalls in the interpretation of probabilistic evidence and show how intuitive reasoning can be misleading in this context. Participants will learn how to critically assess likelihood ratio calculations and to recognise when conclusions are sensitive to modelling assumptions.
Finally, we briefly introduce Bayesian networks as a tool for handling more complex scenarios involving multiple and dependent sources of evidence.
The workshop aims to provide a clear and practically grounded understanding of likelihood ratio-based reasoning, equipping participants to engage critically with forensic statistical evidence.