Show all articles in this section

Let the data speak

AUTHOR: Lachlan Colquhoun   DATE: 30.11.04   ISSUE 3, 2004

Dr. Wood, who joined AGSM last year as Senior Lecturer in Statistics, is a leading researcher in her field and is well known throughout the academic world for her work on non-parametric regression and Bayesian analysis, two theoretical mathematical methods which have been increasingly used in statistical modelling.

{	“Statistical tools give managers competitive advantage.”

ILLUSTRATION: Gregory Baldwin

The worlds of statistical theory and business practice might seem light years away from each other, but Dr. Wood believes business can harness the theories to gain a deeper insight into data, and drive better decision making.

And after an academic career spent in Departments of Statistics in Australian and American universities, she admits that her contact with ‘real’ business data at AGSM has given her a new perspective on statistical modelling.

“I was a bit dubious about working in a business school, I was unsure of what I could offer the business school and vice versa. But talking with the faculty and students about real business problems often gives me insights into statistical theory and also shows me that many of the newly developed statistical methods have a lot to offer the business fraternity,” she says.

“When I taught my first course at AGSM last year the first question students asked was would they use these theories in business, but I have to confess it’s not something I had thought a lot about until arriving here.”

“On reflection I would say that anticipating the future is a major part of any management role; it is the basis of many strategies. Being able to predict more accurately than a competitor is a huge competitive advantage, therefore changes in statistics that improve our ability to forecast or to predict likely outcomes are sources of competitive advantage. So the relevance for business is that it helps create new tools for being able to forecast the future, placing you in a much better position than your competitors, and that’s the bottom line.

When asked what advice she would give to her students, Dr Wood replied “Always look at the data. Often we propose a model for data without checking to see if the assumptions implicit in that model are valid. For example in estimating a linear regression model we assume that the relationship between a predictor and an outcome is linear, but is this an accurate assumption?”

In many cases, people using statistical models employ a ‘black box’ approach, where they simply input data, press a button and let the computer program do the rest, without questioning the assumptions inherent in the model that gives the results.

“One of my messages to students is please don’t just use the pop down menus from some software package and give me a result – I want students to always look at the data and think about what they are doing,” says Dr. Wood.

“When MBA students graduate or progress in their careers they will not necessarily be the ones pressing the buttons. Their job is to ask questions such as “what assumptions did you use in estimating this model?” and “how did you check that your model is valid?” This is very important because as a manager, the inference you draw from a model depends upon the validity of your assumptions.”

Dr. Wood says one of her aims at AGSM is to help educate the Australian business community about how they might apply new analytical techniques.

“Over the last 20 years we have seen an explosion in the availability of data and computer power. This has meant that modelling problems that were previously thought to be intractable have now become feasible and this availability of computer power and data has given rise to many areas of statistics.”

One of these areas is Bayesian Statistics – an approach named after the Reverend Thomas Bayes, 1761. This field of statistics has become increasingly relevant in practical applications because of its ability to estimate flexible models for complex data, she says. Many professional groups such as engineers and meteorologists use Bayesian analysis to model complex data. But Dr Wood feels that Bayesian statistics need not be confined to the more scientific professions. Professions associated with the social sciences, could benefit enormously, she says.

{	“AGSM places a very high value on research, which is unique for Australian business schools.”

PHOTO: Greg Newington (Dr. Sally Wood)

“Bayesian statistics is fundamentally different from the traditional view of statistics that most managers have learned in the past” she says “and by comparison to other professions Australian business has been quite slow on the uptake. It’s a mistake to think you have to be mathematically minded to use Bayesian Statistics, even doctors are beginning to use it and I don’t think of them as particularly numerate.”

Areas such as finance, marketing and organizational behaviour, she says, present opportunities to use contemporary statistical techniques including Bayesian analysis.

“My own research interests lie in non-parametric regression. In non-parametric regression we estimate the relationship between a predictor and an outcome in a flexible manner. By flexible I mean we do not assume that the relationship has any known form and instead let the data speak for itself.”

“In the past when data was scarce and computer power limited or non-existent, restricting a model to be linear made sense. However in today’s world we can afford to relax these assumptions and model relationships between predictors and outcomes less restrictively. A Bayesian approach provides a framework which lends itself to non-parametric regression,” she says.

It’s an approach used to model outcomes as diverse as the safety of medical devices, the time and size of potential terror attacks, and the outcomes of childhood epilepsy. Dr. Wood herself has used them in predicting the likelihood of diabetes occurring in Native American Indian women and in modelling global climate changes.

“One challenge in business is how to harness the new statistical techniques which have been developed over the last 20 years. I’m not saying that traditional methods, such as linear regression should fly out the window – I think they are always your first port of call but they are not the only port of call.”

Dr. Wood says AGSM is unusual for a business school because its statistics group consists of academics who are all internationally respected researchers. “AGSM places a very high value on research, which is unique in Australia,” she says. “In comparison, all of the top US business schools support high levels of research.”

“The knowledge business is like a supply chain. At the upstream end you have researchers, people like myself who develop theoretical models. The aim for the downstream end is to see these theories find a home in business processes and decision making. Clearly there is a time lag between the creation and dissemination of new ideas and techniques. But it is obvious that there are relevant business applications for techniques such as non-parametric regression,” says Dr. Wood.

“In meteorology, for example, they don’t use linear models and business data seems to me to be every bit as complex and every bit as rich as the data they get in meteorology.

“And if they’ve let go of these basic models to forecast the weather then I think its time to let go of them in business.”