In search of a perfect model

As the sophisticated statistical models he builds are directed at questions about climate change and energy demand, Professor Jiti Gao's econometrics research gains a focus on real-life issues that he finds appealing.

The challenge of finding, creating or finetuning the best models to use in analysing often highly complex data provides a satisfying level of intellectual stimulation for Jiti, an Australian Professorial Fellow and an internationally recognised expert in the fields of non- and semi-parametric econometrics as well as time-series and panel data econometrics.

'I was initially trained in the statistical side of econometrics,' Jiti says. 'Gradually my projects have become more motivated by practical issues.'

His interest in the theoretical side of econometrics grew from his training in applied mathematics, and he subsequently developed an interest in both methodology and the empirical application of his research.

Part of Jiti's work involves an initial determination of the most appropriate class of models to apply to data that may come from disciplines as disparate as the social sciences and engineering.

He has a particular interest in time series - data that is measured at successive times, such as the temperatures recorded by meteorological bureaus over the years and used in forecasting or the analysis of trends. They are relevant to collaborative work Jiti undertakes with CSIRO, using econometric modelling to analyse temperature and rainfall series.

Climate studies and econometrics have been closely linked in recent years, he says, because some econometric models are useful for dealing with the large datasets in climatology.

Several Australian Research Council Discovery Project grants support Jiti's research into new methodologies in time series and panel data econometrics. He also has some international funding, notably from China, Norway and Singapore.

In one recent research project that has had a particularly positive impact, he developed a new approach to continuous-time financial models, which among other things are used to improve forecasting of financial returns. Jiti's new model estimation and specification methods mean that financial data may be allowed to 'speak for themselves' in terms of choosing the best possible model from those already available.

In his wider research, says Jiti, the search for a model does not always culminate in a perfect solution, particularly in economics and finance.

'Sometimes there is no best model,' he says. 'You may only be able to find the best approximation.'

It may be the case that several different models are needed for the same data, depending on the requirements. A model that is suitable for forecasting, for example, might not be the best for other aspects.

'You try to get the best model to make sure that you have a relatively accurate modelling procedure to provide solutions to the individual problems,' Jiti says.

Currently, Jiti is coordinating an International Network on High-Dimensional Dynamic Systems supported financially by the Monash Business School. The Network links a group of leading researchers in the Department of Econometrics and Business Statistics at the Monash Business School with a select group of internationally renowned econometricians and statisticians from Cambridge, Columbia, National Tsinghua in Taiwan, Paris, Warwick and Yale Universities, to collaborate on the analysis and application of complex, dynamic and flexible models for high-dimensional statistical data. The establishment of this Network further facilitates and enhances collaborative research activities within the Monash Business School with some of the best and leading centres around the world, such as the Cowles Foundation for Research in Economics at Yale University, Climate Econometrics at the University of Oxford, and the Centre for Research in Econometric Analysis of Time Series at Aarhus University. 

The main aims of the Network are: a) to tackle new and challenging issues in modern econometrics and statistics; b) to use the recognized expertise to establish new research programs in big data analytics; and c) to offer solutions to solve pressing problems in some emerging areas, such as climate and environment, health, ageing and insurance, and stability of financial systems.