The international network is made up of three clusters.
Model building and estimation for high-dimensional time series and panel data
This cluster continues to work both actively and productively towards providing solutions to problems that are at the frontiers of flexible time series and panel data models.
Existing research activities involve:
- climate, energy, health, development and trade data analysis and modelling
- nonlinear methods for nonstationary time series models and predictive regression
- nonlinear and quantile methods for large panel data models
- econometrics for macroeconomics, finance, credit risk modelling, and networking
- high-dimensional econometrics using random matrix theory.
We hosted a two-day, international workshop on High-dimensional time series and panel data in November 2018.
Dynamic high-dimensional data analysis and modelling
Very high frequency high-dimensional time series can be thought of as a stream of data, and methods are required for handling the data efficiently and which update as new data arrive. Typical examples are security sensors which monitor movement or activity every few seconds in a large number of locations.
Researchers at the Monash Business School have developed new algorithms for coherently forecasting such large collections of time series. They are also developing new methods for visualising high-dimensional temporal data, including user interaction and statistical inference.
New inferential and computational techniques for complex statistical models
Researchers in this cluster, in collaboration with leading international statisticians and econometricians, have worked at the forefront of modern Bayesian methods and other computationally-intensive statistical techniques such as indirect inference and bootstrapping.
The research agenda of the cluster has provided rigorous validation of techniques such as approximate Bayesian computation, indirect inference and the non-parametric bootstrap, as well as formulating new ways of applying such techniques to challenging models. Probabilistic forecasts, which provide a complete representation of the uncertainty associated with future outcomes, have been a focus of this cluster.