2013 Australian Statistical Conference

Celebrating 50 years, the Statistical Society of Australia held this year’s conference in Adelaide. This premier event sees world class Australian and international statisticians come together to share knowledge and expertise. Data Analysis Australia continued its support of the statistics community by sending a contingent of six delegates to the conference, each of whom gave a presentation. The presentations covered topics drawn from the large variety of projects worked on recently at Data Analysis Australia. On the lighter side Data Analysis Australia sponsored the Young Statisticians’ Dinner as part of our ongoing support of the profession at all levels. 

Our talks highlighted the diverse range of industries in which Data Analysis Australia has expertise:

John Henstridge spoke on analysis we had carried out for the mining sector, presenting an interesting overview on the interplay between mineral processing and the spatial aspects of the mine and how this leads to some statistical issues that are rarely considered. Data Analysis Australia has assisted with the analysis of data from a single site, in particular, quantifying the biases. These bias corrections have assisted our client with decision making about the likely recovery rate of minerals at a particular mine site. 

Taryn Major spoke about how some of our clients require model-based, automated decision-making on a daily, hourly or even half-hourly basis. Taryn’s presentation looked at some of the issues associated with automated model fitting as it relates to day-ahead electricity forecasting. In Western Australia, electricity demand is strongly associated with the weather and therefore an electricity demand model must be robust and provide good forecasts in both summer and winter. Data Analysis Australia consultants have provided automated solutions which generate quality forecasts on an ongoing basis to several clients.

Prudence Thompson gave a presentation on some of the challenges Data Analysis Australia has experienced when trying to quantify the impact of behaviour change programs. One example of these challenges included the definition of a measurement unit for recidivism - if a drug addict stops using drugs but replaces this with alcohol abuse, are they really rehabilitated? Another challenge was the existence of confounding factors such as simultaneous intervention programs. Data Analysis Australia has quantified the impact of several behaviour change programs, tailoring our analysis to deal with the challenges presented by each program. 

Emma Smith described a Port simulation model developed by Data Analysis Australia. The simulation model was used to provide strategic information for the Port in a relatively risk-free manner by enabling the possible impact of various available options to be assessed, with only the best option needing to be implemented. The model included the physical characteristics of the Port, such as channels and berths, as well as the human resources required to facilitate vessel travel and berthing. Such models can be used to inform business decisions. 

Alethea Rea spoke on the challenge of estimating quantities for small geographical areas when the information is available only for large areas such as by state. She presented a technique known as small area estimation and its application to estimating the distribution of healthcare card holders across the Perth Metropolitan area. 

Kathy Haskard joined Data Analysis Australia from the public sector, and she found some interesting differences in the way Data Analysis Australia operated, both as a private company and as the largest group of co-located statisticians she has worked with. Kathy’s presentation highlighted the benefits of working in teams for both staff, who learn from each other, and clients, who benefit from a team specifically tailored to each project.

All our talks were well received and those of us who attended the conference had an enjoyable experience. We look forward to the 2014 conference. 

September 2013