Efficient Design for Generalized Linear Models. In industrial, commercial and social research, we collect data in order to predict the outcome of a process based on the inputs to that process. We want to maximize the information that is gained from the data. Good planning is crucially important to achieve this. This project will determine how best to select the inputs to the process for many situations that occur in research. A computer package to answer these questions will be written. The nati ....Efficient Design for Generalized Linear Models. In industrial, commercial and social research, we collect data in order to predict the outcome of a process based on the inputs to that process. We want to maximize the information that is gained from the data. Good planning is crucially important to achieve this. This project will determine how best to select the inputs to the process for many situations that occur in research. A computer package to answer these questions will be written. The nation will benefit from a fundamental increase in efficiency of research and, therefore, in efficient use of research dollars.Read moreRead less
Trans-dimensional and Approximate Bayesian Computation. Many applied scientists in Australia, particularly those in the biological, medical and environmental sciences are now interested in incorporating Bayesian statistical methodologies into their research.
The development of more generic and efficient Bayesian statistical methods will not only benefit applied statisticians but also the more occasional users of statistics in other disciplinary areas. The success of this project will enhance Au ....Trans-dimensional and Approximate Bayesian Computation. Many applied scientists in Australia, particularly those in the biological, medical and environmental sciences are now interested in incorporating Bayesian statistical methodologies into their research.
The development of more generic and efficient Bayesian statistical methods will not only benefit applied statisticians but also the more occasional users of statistics in other disciplinary areas. The success of this project will enhance Australia's reputation as a strong contributor to the development of Bayesian methodologies. Two PhD students will also be provided training in computational Bayesian statistics.Read moreRead less
New Bayesian methodology for understanding complex systems using hidden Markov models and expert opinion, environmental, robotics and genomics applications. This project aims to merge four areas of intense international interest in describing complex systems: hidden Markov models and mixtures, semi-parametric and nonparametric approaches, true combination of expert opinion with data, and new Bayesian computational methods based on perfect sampling and particle sampling. The project will signific ....New Bayesian methodology for understanding complex systems using hidden Markov models and expert opinion, environmental, robotics and genomics applications. This project aims to merge four areas of intense international interest in describing complex systems: hidden Markov models and mixtures, semi-parametric and nonparametric approaches, true combination of expert opinion with data, and new Bayesian computational methods based on perfect sampling and particle sampling. The project will significantly contribute to statistical methodology and its ability to inform about real-world problems. A strong focus on applications to genomics, robotics and environmental modelling will bring immediate research and monetary benefit for industry. Expected outcomes include enhanced cross-disciplinary and international linkages, publications, industry-funded projects and highly trained graduates.Read moreRead less
Uncertainty, Risk and Related Concepts in Machine Learning. Machine learning is the science of making sense of data. It does not and cannot remove all risk and uncertainty. This project proposes to study the foundations of how machine learning uses, represents and communicates risk and uncertainty. It aims to do so by finding new theoretical connections between diverse notions that have arisen in allied disciplines. These include risk, uncertainty, scoring rules and loss functions, divergences, ....Uncertainty, Risk and Related Concepts in Machine Learning. Machine learning is the science of making sense of data. It does not and cannot remove all risk and uncertainty. This project proposes to study the foundations of how machine learning uses, represents and communicates risk and uncertainty. It aims to do so by finding new theoretical connections between diverse notions that have arisen in allied disciplines. These include risk, uncertainty, scoring rules and loss functions, divergences, statistics and different ways of aggregating information. By building a more complete theoretical map it is expected that new machine learning methods will be developed, but more importantly that machine learning will be able to be better integrated into larger socio-technical systems.Read moreRead less
New approaches to predictive modelling of high-dimensional count data to study climate impacts on ecological communities. This project will lay methodological foundations for future studies of potential impacts of climate change on ecological communities. A flexible new toolset of predictive modelling approaches will be developed, capable of handling all common data types, which fit easy-to-interpret models, and which are more powerful than currently used methods.
Principled statistical methods for high-dimensional correlation networks. This project aims to develop a novel and principled approach for building correlation networks. Correlation networks aim to identify the most significant associations present in modern massive datasets, and have numerous applications, ranging from the biomedical and environmental sciences to the social sciences. Nodes of such networks represent features, and edges represent associations, or the lack thereof. Current method ....Principled statistical methods for high-dimensional correlation networks. This project aims to develop a novel and principled approach for building correlation networks. Correlation networks aim to identify the most significant associations present in modern massive datasets, and have numerous applications, ranging from the biomedical and environmental sciences to the social sciences. Nodes of such networks represent features, and edges represent associations, or the lack thereof. Current methods are not readily scalable to modern ultra-high dimensional settings, and do not account for uncertainty in the estimated associations. This project will develop a principled, highly scalable methodology for building such networks, which incorporates uncertainty quantification. Emphasis is placed on modern ultra-high dimensional settings in which differentiating a true correlation from a spurious one is a notoriously difficult task.Read moreRead less
Bayesian inference for complex regression models using mixtures. The project will use mixtures to flexibly model complex regression functions and will develop Bayesian methods for carrying out statistical inference on these models. The models will deal with both Gaussian and non-Gaussian data. Multiple explanatory variables are dealt with by mixing simple additives to produce flexible high dimensional function estimates. Variable selection and model averaging will be used to identify important v ....Bayesian inference for complex regression models using mixtures. The project will use mixtures to flexibly model complex regression functions and will develop Bayesian methods for carrying out statistical inference on these models. The models will deal with both Gaussian and non-Gaussian data. Multiple explanatory variables are dealt with by mixing simple additives to produce flexible high dimensional function estimates. Variable selection and model averaging will be used to identify important variables and thus make the estimation more efficient. The methods will be extended to multivariate responses where account will taken be taken of the structure of the dependence between responses.Read moreRead less
New methods for modelling real-world extremes. This project aims to develop new theory and methods for analysing and predicting extreme values observed in real-world processes. Many existing techniques are limited by convenient mathematical assumptions that commonly do not hold in practice: dependence at asymptotic levels, process stationarity, and that the observed data are direct measurements of the process of interest. As a result, using these techniques may produce undesirable results. Expec ....New methods for modelling real-world extremes. This project aims to develop new theory and methods for analysing and predicting extreme values observed in real-world processes. Many existing techniques are limited by convenient mathematical assumptions that commonly do not hold in practice: dependence at asymptotic levels, process stationarity, and that the observed data are direct measurements of the process of interest. As a result, using these techniques may produce undesirable results. Expected outcomes of this project include theoretically justified data analysis techniques that can accurately model extreme values seen in the real world. Project benefits include more realistic analyses of nationally important applications in climate, bushfire insurance risk, and anomaly detection.Read moreRead less
Building models for complex data. The purpose of this project is to better understand the process of building statistical models and construct new methods for building models for particular kinds of complex data. The expected outcomes include a new way of thinking about model building and practical tools which together enable us to get more value out of analysing complex data.
Semiparametric Regression for Streaming Data. Semiparametric regression converts large and complex data-sets into interpretable summaries from which sound decisions can be made. This project tackles semiparametric regression analysis of streaming data - where the data are so voluminous that they may not be storable in standard computer memory and therefore need to be processed rapidly on arrival and then discarded. Effective solutions necessitate a rethinking of semi-parametric regression and ne ....Semiparametric Regression for Streaming Data. Semiparametric regression converts large and complex data-sets into interpretable summaries from which sound decisions can be made. This project tackles semiparametric regression analysis of streaming data - where the data are so voluminous that they may not be storable in standard computer memory and therefore need to be processed rapidly on arrival and then discarded. Effective solutions necessitate a rethinking of semi-parametric regression and new approaches will be developed. The project will also develop novel theory and methodology for robotics applications. It will allow analysis of streaming and massive data sets that would not be possible using currently available methods, opening up new applications.Read moreRead less