Quantitative modelling of complex systems is integral to all areas of business, industry and government in which decisions are made on the basis of data. Improved methods and models for understanding, predicting and improving systems in areas such as Health, Science, Finance, Engineering and the Environment are of central, national importance.
Research in statistics is organised through the Statistical Modelling and Inference research group and a number of the group are participants in the Centre for Complex Dynamic Systems and Control.
The statistics staff within the School of Mathematical and Physical Sciences offers a team of researchers in the following key areas:
Assessing statistical models using smooth tests of goodness of fit, constructing informative analyses of important experimental designs making minimal assumptions - using contingency tables to construct informative nonparametric analyses.
- Study of Distributions
Behaviour estimation and application of flexible distributions such as the g-and-k and generalised lambda distribution. Data fitting and modelling with such distributions.
- Time Series
Application and understanding of linear and nonlinear time series models, including Smooth Threshold Autoregressive models, Ergodic theory, Bayesian approaches, forecasting dynamic models, and the application of time series models to MCMC convergence and hidden Markov models.
- Bayesian Modelling
Construction of hierarchical models. Development and assessment of numerical (principally Markov chain Monte Carlo) methods for analysis. Model comparison and adequacy, and variable selection, modelling on bounded spaces.
- Data Mining
The integration and application of data mining methods such as recursive partitioning and neural networks for modelling large and complex datasets.
- Ecological and Environmental Modelling
Modelling rare events, combination of different sources of information, assessing ecological models, methods for forestry and remote sensing.
Risk stratification, control charts, novel methods for modelling and prediction, causation, meta-analysis and Bayesian modelling. Quality improvement and Bayesian hierarchical modelling. Modelling using GAMs, GLMs, logistic regression and structural equations.
- Process Improvement
Methods for and application of process improvement. Total quality management approaches, statistical tools for quality, statistical data analysis, design of experiments, sampling, methods for evaluating change and risk, evolutionary operations and development of appropriate control charts. Operations research including linear programming, inventory management, optimisation and scheduling.
- Performance Measurement
Analysis, reporting and management.
- Statistical Education
Research into, and application of, best practice in teaching and learning statistics.
Statistical approaches to machine learning, particularly classification using mixtures for automatic colour recognition, face recognition and Bayesian belief updating for robot localisation. Research through the Newcastle Robotics Laboratory.
- Socio-Spatial statistics
Bringing together multi-variate and spatial statistical methods to characterise social structure, such as characterising disadvantage in a particular area. This joint research with the Centre for Urban and Regional Studies and the NSW government and has also led to work on the processes required for data sharing with government agencies.
Statistics Working Papers
Number 2008/1 "On building an academic research profile" John Rayner
The Discipline holds regular seminars relating to the work of the above research areas. For information about our seminar program please see the Statistics Research Seminars page.