Image

Presentation: A Clinically Meaningful, Easy-to-Interpret, Effect-Size

Presenter: Emeritus Professor Michael Smithson is part of the School of Medicine and Psychology at The Australian National University and a Fellow of the Academy of Social Sciences in Australia. He has contributed to research and theory on judgment and decision-making under uncertainty and ignorance, statistical methods for the social sciences, and applications of fuzzy set theory to the social sciences. He is the author of 7 books and co-editor of 2 books, and his other publications include more than 200 refereed journal articles and book chapters.

Although we all agree that effect-sizes are important, they often don't tell us what we would like to know. For instance, suppose a controlled clinical study comparing depression treatments A and B reports that mean post-treatment BDI II scores are significantly lower for recipients of A than for B (p 0.001, Cohen's d = -0.8).  How likely is it that someone given treatment A will end up less depressed than if they’re given B instead?  Cohen’s d doesn’t answer this question, but I’ll be presenting an effect-size that does.  This effect-size can be applied to ordinal and quantitative dependent variables in any GLM and is base-rate insensitive, robust to outliers, and invariant under order-preserving transformations.  Moreover, it already exists in most statistical software environments—but is well-disguised. My talk will remove the disguise and demonstrate its applications in real data-sets.

In-person attendance is strongly encouraged. There is a Zoom option for those unable to attend in-person.

Tea/coffee and biscuits will be provided after the seminar. We encourage you to bring along a mug to help us minimise waste.