Discriminability in deception is not d: Reporting the Overlap Coefficient for practitioner-accessible audiences

Applied psychology aims to develop evidence-based conversation between researchers and practitioners. We should aim for these conversations to be more transparent and accessible, including in terms of how we summarise and discuss statistical analysis. However, classically deployed mean-difference statistics can hide shared variance between conditions and do not truly reflect researchers’ aims of ‘differentiating’ or ‘discriminating’ conditions. Importantly, mean differences do not provide practitioners with meaningful guidance on how to interpret one case at one point in time. Here, through focusing on deception detection research I provide an introduction to using the overlap coefficient (OVL) to enhance research-practice conversations. I highlight that even large mean differences (d= 3.00) can have one in ten cases presenting ambiguously (OVL= 0.13). I argue that reporting the overlap (and non-overlap) values and framing our results in terms of ‘percentage of cases differentiated’, allows us to better communicate our findings to practitioners. The use of the OVL statistic allows us to temper and expand the reporting of findings in applied psychology and will enhance practitioner-research communication.