Our work

Data integration and complex modelling

We augmented citizen science counts of 326 bird species with weather, geospatial and ecological information and used generalised additive mixed-effects modelling to examine seasonal effects and long-term trends in species’ presence and abundance. A focus was the role of noisy miners in displacing other bird species.

Meta-analysis and expert witness testimony

Our meta-analyses of evidence from 438 studies of stress urinary incontinence and 130 studies of pelvic organ prolapse supported Professor Ian Gordon’s expert witness testimony in a class action over transvaginal mesh implants. The judge extensively cited his statistical and epidemiological evidence.

Julie Davis, speaking outside court. Image via AAP.

Forecasting demand

A large, complex and innovative data analytics project integrated different data sources to provide Victoria Legal Aid with forecasts of the level of demand for legal services based on the expected growth in the Victorian population.

Expert review and survey methodology

Collaborating with the Melbourne Energy Institute, our experts reviewed methodology, advised on surveys and sampling, assessed reports and supervised a major study on “revealed preferences” to support the Australian Energy Regulator’s need to estimate what customers are willing to pay to secure more reliable electricity.

Machine learning

We used machine-learning survival analysis modelling to understand prognosis in advanced lung cancer, and compared models using: (a) clinical data, (b) radiomics data, or (c) both radiomics and clinical data. The tree-based models allowed us to include over 1000 radiomics features.

Modelling in the presence of missing data

We supported a graduate researcher in the use of inverse probability weighted adjustment to account for missing data in hospital reports of antimicrobial-resistant pathogen incidence. This allowed for the estimation of state-wide incidence of such pathogens in Victoria over a 35-year period.

Sampling design

We have devised sampling schemes for agencies holding valuable public collections to enable sound valuations or counts of the collection to be made, along with estimates of uncertainty. Agencies include Arts Centre Melbourne, The Australian Museum, and Australian Centre for the Moving Image.

Analysis of experiments

Our analysis and meta-analysis of field trials, both experimental and commercial, provided evidence supporting the effectiveness of the organic liquid seaweed fertiliser Seasol in promoting growth in a range of crops, including grapevines, sugar cane, and avocado, and providing a foundation for ongoing research.

Experimental design

We designed experiments for the Department of Agriculture to evaluate 3D x-ray technology for screening passenger luggage to identify incoming biosecurity risks, and used outcomes of analysis of these experiments to inform field trials and additional experiments of related technologies.

Sample design, precision and sample size

Design of a compositional waste audit for Sustainability Victoria required the estimates of the composition of waste (plastic, glass, food, etc.) generated by various business types to be as precise as possible. We assessed the expected margin of error given the audit plan.

Evaluating AI

A randomised controlled trial is underway to assess if the implementation of an AI mammogram reader improves breast cancer screening.  We provided critical input into statistical aspects of the trial design, especially in determining sample sizes for this very large study.

Cluster randomised trial design

We provided design advice and performed the randomisation of treatments to clusters for a feasibility trial of a care bundle including nurse training, designed to improve indwelling catheter care in aged care homes (the clusters).