Statistician Uncovers New Method for Analysing Geochemical Data

Dr Janice Scealy, a lecturer with the ANU Research School of Finance, Actuarial Studies and Statistics has co-authored a paper that has been published in the Journal of the American Statistical Association.

The paper, Robust Principal Component Analysis for Power Transformed Compositional Data proposes a new way of analysing sediment or rock samples collected from geographically dispersed sites. 

Working with Geoscience Australia on a set of data from sediment samples from sites across Australia, Janice’s proposed method is able to identify geochemical patterns more easily and relate these to known geological processes. 

“One of the major goals in the analysis of their data is to help scientists link the samples collected to known geological processes such as residual weathering phases, carbonate formation, and so on,” said Janice.

“Principal Component Analysis is a well-established method and now we’re getting a huge improvement, so hopefully people will pick this up and they’ll start using our method in the future,” she said.

Janice did her PhD in analysis methods for proportions data, so this work with Geoscience Australia builds on her existing research and modifies it.

In the long term, she hopes to eventually build an R package - a statistical software package that other statisticians could download and use to repeat the same analysis, improving the way geochemical survey data is analysed.

“Publishing in The Journal of American Statistical Association, one of the top three statistics journals, is an excellent achievement,” said Professor Steven Roberts, Director of the ANU Research School of Finance, Actuarial Studies and Statistics.

“The collaborative nature of this research which included Geosciences Australia and the Geological Survey of Canada is also noteworthy and attests to the substantive problem in the analysis of geochemical survey data addressed by the research,” he said.