- This event has passed.
National Institute for Applied Statistics Research Australia Seminar Series
February 12 @ 2:30 pm - 3:30 pm
Professor Irene Hudson (Swinburne University of Technology)
Multidisciplinary research with impact: the integration of statistics and data science in research
Mapping, modelling and visualization of micro systems (molecules, wood anatomy, brain voxels) and of macro systems (global tobacco use with the WHO, climate change, risk/prevalence estimation) underpin my collaborative research, along with computational intelligence. My academic research has been aligned with active consulting with the forest and pulp and paper industry, the biomedical health sector industry drug discovery and the railways industry. I currently collaborate with the following external research organisations developing statistical methods for:  the WHO Tobacco Free Initiative (TFI) on global health and tobacco control (Lancet 2015);  Imperial College on Bayesian Dirichlet Process Mixture Models (DPMMs) to accommodate high dimensional trajectory data extended to incorporate survival endpoints (submitted to Statistics in Medicine, 2018);  CqU’s Appleton Institute for Behavioural Science, Human Factors and Safety, on modelling work/rest/sleep scheduling to minimise fatigue in the Australian railway industry (Hudson, et al., 2016, IEEE 2016, BDVA);  the ARC Centre of Excellence for Nanoscale BioPhotonics (CNBP) on image cytometry and data mining;  the Hunter Medical Research Institute (HMRI) and Engineering on stroke rehab devices and modelling functional states (this research was funded by an NHMRC DG 2015-2017, ARM MOVEMENT MEASURE (ArMM) FOR STROKE RECOVERY, UoN Innovation award, 2015). I likewise engage in social research – with the NSW Bureau of Crime Statistics, Statistics NZ on official stats/voting patterns and the ABS on surveillance, aim of the proposed surveillance research is to align within a new framework capture –recapture methods with ecological inference and Categorical data methods developed by Eric Beh, Hudson et al (Hudson, Moore, Beh, Steel, DG. (2010), JRSS A; Beh, Tran & Hudson (2013) Computational Stats & Data Analysis, and Tran, Beh, Hudson (2018), Stat Journal of IAOS, Beh et al., (2015) Chapter in Advances in Latent Variables, Springer.
I shall discuss novel research methods development to model state systems (e.g. sleep states) via 2 approaches: Dirichlet Process Mixture Models (DPMMs) to model sleep, mortality and morbidity in a cohort of very aged women the Australian Longitudinal Study on Women’s Health (www.alswh.org.au) (Leigh, Hudson, et al., 2015, 2016 and 2018);  A multivariate extension of the mixture transition distribution (MTD) model, to accommodate covariate interactions to study Australian railway drivers (RDs) sleep patterns by creating RD networks of sleep/wake /duty/break feature parameter vectors of between-states transition probabilities (Hudson et al, 2016, chapter in Computational Intelligence; Springer).
Further discussion will focus on the integration of data mining, CA and mixture analytic methods in the following areas: Chemo-informatics: work with Adelaide University (ARC DP) to correlate the molecular structure and true binding IC50 of potential small molecules for the treatment of cataracts (Hudson et al. 2016, Computational Intelligence), and the analysis of the DrugBank database, to map and visualise chemo-space. Developed new druggability rules and shown that high scoring ligands are associated with particular disease targets (Hudson et al., 2017), beyond Lipinski’s Ro5 for the drug industry. New work commencing on Protein-Peptide interactions (in prep for PLoS ONE).  Imaging cytometry with the CNBP: to extend methods developed for high dimensional, high throughput scenarios in cytometry, the derivation of quantifiers of features, from the measurement of large populations of single cells. The focus of the project is automated imaging flow cytometry (IFC), which integrates flow cytometry with digital microscopy to produce high-resolution digital imaging with quantitative analysis. To date, data analytic methods for imaging cytometry have not been able to gain full leverage of the rich information in digital imagery, now seen as a critical enabling and driving technology in molecular and cellular biology, biotechnology and medicine. Currently there is keen interest to apply imaging cytometry for diagnostic assessment of diseases, but no instrument exists to date to accomplish this in routine diagnostic laboratories, as such the research has significant potential impact. This work involves a unique biological microscopy (of > 260,000 white blood cells).
Other areas of theoretical and application methods are: Causal inference, Copula methods, Data mining of high dimensional time series and images, Climate change, Sports analytics, Trajectories in hand motion, choice & cognition (applicable to devices and computer science software engineering), Markovian systems, Math modelling; ICU metrics of agitation-sedation control, Bayesian Meta analysis, Engineering Devices, and Neuroscience.