Copyright © 2012
Data Analysis Australia

Newsletter Articles

Data Analysis Australia includes expanded versions of our newsletter articles online.

Featured Articles

 Survey Best Practice - Why Bigger Is Not Always Better

Surveys can be a simple do it yourself satisfaction survey, an in-depth longitudinal study that can go for years or a closely followed political poll.  They can be implemented by telephone, face to face or the internet and only your imagination limits the topics that can be surveyed.  They cannot be ignored and are an integral source of information for researchers.  They can drive change from government policy to the label on a shampoo bottle.  Although surveys may differ greatly in design and purpose, if they do not adhere to good practice in questionnaire design, sampling and weighting, the results can be very poor, even wrong.   The trade-off between length of questionnaire, method of delivery, sampling size, complexity of analysis and cost needs a level of expertise.  This month’s article examines some of the issues of survey design through the past turmoil’s of political polls as they endeavour to predict the outcomes of US presidential elections.

Australian Statistical Conference

Celebrating 50 years was the theme of this year’s Australian Statistical Conference held in Adelaide.  International and Australian delegates came together to develop and share their knowledge and expertise.  Data Analysis Australia helped the society celebrate by sending six staff to Adelaide, all of whom presented on a wide range of topics.  The team highlighted the diversity of projects carried out by Data Analysis Australia.  Presentations included recent applications of statistical techniques to mineral processing in the mining sector, modelling a port operation, and the electricity market.  Other talks highlighted our extensive work in analysing behaviour change programs and the challenges and uniqueness of working in a large commercial statistical company.  All our talks were well received and those of us who attended the conference had an enjoyable experience meeting other attendees and visiting Adelaide and its surrounds.  

Role of the Statistician

The needs of the twentieth century have required a change in the role of the statistician.   Previously often seen in a support role to science, statistics has now become an important tool not only to science but to technology and business.  Large amounts of data combined with computing power have led to statistical innovation and breakthroughs such as those seen in the genome-wide studies associated with diseases.  Over the years, Data Analysis Australia has worked on a variety of projects relating to the Justice system - Perth has a fast growing population and forecasting the demand for facilities such as court complexes is vitally important.  The issues are many, involving demographic projections, an understanding of trends and spatial variation in crime, how courts process matters, changes in legal procedures and the need to understand the planning decisions that the forecasts are used for.  John Henstridge discusses the role of the statistician as a leader in this context, as well as the various roles we as statisticians have had across all our projects, from the initial stages of design and evaluation to analysis and interpretation and beyond.  

Making Sense of Complex Data

Large data sets are becoming more common place as technology allows us to more easily collect and store information.  But where the analysis becomes more interesting is not so much in the number of records as in the number of variables.  The analysis of this complex data requires the reduction of the dimensions without reducing the information within the data.  A popular method developed by Karl Pearson (1901) is Principal Component Analysis (PCA), used to reduce a complex data set to a lower dimension, giving a simplified structure which can be more easily interpreted.  On the front page of the March newsletter, the discussion revolves around complex data and how PCA can be a useful methodological tool in understanding data with large numbers of variables. 

The importance of Good Data

Reliable, meaningful results are the aim of any data analysis and good analysis relies on good data.  For this we need to go back to the beginning.  Data Analysis Australia believes that the starting point is in the planning of the data collection itself.  Data is a resource often limited by constraints of timing, budget and source.  Statisticians have the expertise to use the right methodologies for every client's unique requirements.  Methodologies such as sampling theory and experimental design can be used to maximise the amount of useful information and increase efficiency.  In this month's newsletter, John Henstridge gives a snapshot of how planning the data collection itself is the important first step in ensuring meaningful results from any analysis.

Change is inevitable, but is it real?

 In the statistical world identifying change at a known time is a standard task.  However, identifying sequences of change over time where the timing of changes is unknown is a far more complex problem.  The combination of advances in statistical and mathematical theory has made these tasks feasible.  The final ingredient is the expertise of Data Analysis Australia to use the right methodology in modelling the data.  Recently, we used 'breakpoint' detection methods to determine change in data.  One example shows where the breakpoint method was used to check whether laboratory test results were consistent where expected, or whether they were affected by uncontrolled factors over time.  The other example was a question of just how much historical data reflected current conditions in order to appropriately identify the current operating trends for a utility provider, and to use those trends to produce forecasts.  Emma Smith and John Dickson have been actively working in this area of breakpoint analysis, which is discussed further in the leading article of the September newsletter. 

Surveys and Technology

The increasing popularity of mobile phones and other new technologies replacing traditional landline phones is creating new challenges in the sampling and data collection for surveys.  Recent research suggests differences in the mobile only population compared to other populations, meaning that for some key surveys, excluding mobile phone only households could have disastrous effects.  However, contacting mobile phone numbers for a survey is not straight forward, as the number could be attached to a non phone form of technology (eg. an iPad), the locality (or even time zone) cannot be determined from the phone number and various safety considerations (such as driving) can arise.  Anna Munday and John Henstridge participated in a recent workshop which brought together a number of stakeholders concerned about this issue.  In this article they discuss the effect new technology has on data collection and surveys and how Data Analysis Australia provides solutions to overcome these issues.

How Extreme is Extreme?

We have all seen the effects of extreme weather conditions over the recent months.  Cyclones and associated rain have caused flooding, leaving devastation in their wake.  Can these extremes be predicted or forecasted and what part can statistics play in these 'once in a life time' events?  Extreme value statistics provide a means of planning for these events, and for other important but less dramatic events that affect us every day.  In the main article of the newsletter, John Henstridge talks about the weather and how Data Analysis Australia has applied such methods to a variety of problems from energy demands to financial risks.

Australian Statistical Conference 2010

Data Analysis Australia was heavily involved in the recent Australian Statistical Conference 2010, held in Fremantle.  We were a proud silver sponsor of the event, which proved to be a memorable and exciting week.  All Data Analysis Australia Statistical Consultants attended and contributed in some way, joining over 350 of their fellow delegates from across Australia and around the world.  It was a great opportunity for our Consultants to learn from their industry peers, as well as to showcase some of the exciting and challenging work that we undertake with our clients.  We also enjoyed catching up with old friends, and making some new ones, in the course of the Social Program.  

How Many 'Bad Apples' Are Too Many?

In today's society we demand a high quality in the items we purchase.  Perhaps surprisingly it was the military in WWI that had a major impact on the quality movement.  A demand for high quality in the shells used in trench warfare led to a quality standard being developed.  Since the final test of a shell is to fire it, a sampling regime was required to test batches and decide when to either accept or reject the whole batch.  Today the Standards embody many good ideas but are not flexible for all circumstances.  Data Analysis Australia uses our expertise in sampling and our experience in understanding clients' problems to develop customised solutions.  

What is a Data Analysis Australia Consultant?

The term 'consultant' can be applied to a variety of jobs and the term on its own does not describe the role of the individual.   John Henstridge reflects on his view of how the term 'consultant' should be applied to describe staff roles at Data Analysis Australia - not as a noun, but as an adjective describing the manner in which we practice our technical professions.

Hockey Sticks, Climate and Statistics

Everyone has an opinion on climate change.  How has that opinion come about and what is it based on?  Climate change has become a public debate putting pressure on scientists and politicians alike.  However the main point may well be in the risk of acting or the risk in not acting.  John Henstridge looks at the climate change debate from a fresh angle.  He discusses some of the background data and how statisticians can provide another level of understanding.  

Is the Price Right and How Do You Choose?

How do consumers decide what they want?  Is it brand, colour or price or a complex combination of many things?  Determining the drivers of consumer decision making requires careful consideration of sampling, questionnaire design and modelling.  Much market research is aimed at this field using tools such as contingent value analysis, conjoint analysis, modal choice analysis or trade-off analysis.  At Data Analysis Australia we understand the underlying mathematics behind these concepts and are able to apply the correct modelling procedure to the specific requirements of the client.  It is our combined strength in mathematics and the understanding of market research objectives that gives Data Analysis Australia the edge.

Will Australia Count in the Future?

Although statisticians and mathematicians frequently work 'behind the scenes', many advances that make our modern world modern depend upon their skills.  Just one example is the Human Genome project, where at the core, sophisticated mathematical methods are putting together millions of disparate pieces of information into a coherent whole, speeding up the process a hundredfold.  In this issue we draw attention to the increasing rarity of mathematical and statistical skills in Australia.  This is a matter of concern for everyone.

Shoemakers' Lasts and Model Ships

What do shoemakers' lasts and model ships have in common and how do they relate to Data Analysis Australia? The answer can be found in this month's newsletter article which discusses how it is not just creating a model that makes it useful, but ensuring that it is correct by understanding both its underlying structure and the data that feeds into it. It is this duality of understanding that gives Data Analysis Australia the expertise to present relevant, reliable results and recommendations to our clients. 

Seeing the Unforeseeable

The past six months has seen a meltdown of the global financial system. Many people have been forced to reconsider just what is a good investment and ask "why did no-one see this coming?". These are questions of risk and forecasting, areas where statisticians and mathematicians have a special view. Data Analysis Australia, where possible, calculates the bounds on the certainty of any forecast that we produce - what is most likely and a range of what might happen. We consider a "good" forecast one that gives information to our clients about the risks and the positives.

Celebrating Diversity

"Celebrating Diversity" was the theme of the recent Australian Statistical Conference in Melbourne, where over 300 statisticians attended - the largest gathering of statisticians in Australia.  Data Analysis Australia sent a strong contingent to the conference where we presented a number of highly varied topics demonstrating the diversity of our own work. 

Imagining Numbers

The famous saying, "a picture is worth a thousand words", is paralleled by statisticians who might say that a picture can be worth a thousand numbers.  In fact, every day we are presented in newspapers or on television with graphics trying to get across some form of numerical message, but how accurate are these graphics? 

Past Articles