The South Australian Branch of the Statistical Society would like to invite you to an in-person and online presentation by Prof Dot Dumuid, Aaron Miatke and Dr Ty Stanford.
For those interested there will be a dinner after the presentation.
Date: 6th August (Wednesday) 2025
Time: 6:00 - 7:00pm (Adelaide time). Pre-meeting refreshments and networking between 5:30 – 6:00pm
Venue: In-person and Zoom
Dinner: A dinner will be held after the meeting. Location TBA
Please RSVP for dinner to Lan.Kelly@unisa.edu.au by 4th August (Monday), as we are usually unable to change the booking numbers at the last minute.
Abstract
Compositional data are everywhere! We can find them in chemistry (alloy composition, drug formulation), economics (consumer spending, market shares), environmental (soil composition, air pollution sources), and health research (body composition, dietary intake, microbiome, time use). Compositional data are made of mutually exclusive and exhaustive non-negative components that sum to a constant. Their constant sum constraint introduces a dependency structure between the components, where if you know all but one component, the last is determined. As a result, statistical analysis of compositional data should consider the components relative to each other. Compositional data analysis provides a way to do this by expressing the compositions (absolute information) as a set of log-ratios (relative information) prior to their inclusion in statistical models such as regression. This talk will introduce the principles of compositional data analysis and provide practical examples of its application based on our research within time-use epidemiology.
Speakers
- Speaker 1: Professor Dot Dumuid
- Speaker 2: Dr Ty Stanford
- Speaker 3: Aaron Miatke
Dot Dumuid is a Professor of Behavioural Epidemiology at the University of South Australia. She researches how the time we spend in different daily behaviours impacts our health, with a focus on finding a healthy balance between all the daily competing demands on our time.
Ty Stanford is a biostatistician working at the University of South Australia. He met Professor Dot Dumuid, thanks to a referral from Statistical Society member Andrew Metcalfe, to help on a compositional data analysis (CoDA) problem almost a decade ago. Ever since then he has been drawn to extending and applying statistical CoDA techniques, specifically relating to health measures and 24-hour time-use compositional data.
Aaron Miatke is a final year PhD student at the Alliance for Research in Exercise, Nutrition and Activity at the University of South Australia. His research explores how to model multi-level compositional data with the purpose of evaluating time-use intervention trials and longitudinal cohort studies.
@StatSocAus
