Branch Meetings


E.A. Cornish Memorial Lecture Announcement – South Australian Branch of the Statistical Society of Australia

The South Australian Branch is pleased to announce the 2017 E.A. Cornish Memorial Lecture will be presented by Professor Robert Elliott, Research Professor at the University of South Australia.  Held every two years, this lecture is the highlight of our speaker program.

Date:  Wednesday 25 October 2017
 Napier G04 Lecture Theatre, University of Adelaide, North Terrace. 
:   Pre-meeting drinks & nibbles at 5:30 pm, E. A. Cornish Memorial Lecture at 6:20 pm.

The title of Robert’s lecture is “New Ideas in an Old Framework

The binomial model, based on a discrete time random walk, is a standard framework used to introduce risk neutral pricing of financial assets. Martingale representation, backward stochastic differential equations and the Malliavin calculus are difficult concepts in a continuous time setting. This talk will present these ideas in the simple, discrete time binomial model.





The South Australian Branch of the Statistical Society would like to invite you to the February talk of the 2016 program.


Venue: Ingkarni Wardli B17, North Terrace, Adelaide University.

Campus map is available at


***Please note that most entrance doors to Adelaide University buildings close at 6pm so make sure you arrive in time for the talk.



5.30pm – Refreshments in the Lecture Theatre.

6.05pm – General Meeting Talk.

7.30 pm – A dinner will be held after the meeting at Amalfi Pizzeria Ristorante, 29 Frome Street, Adelaide.

Please rsvp for dinner to [email protected] by 22nd February.


Speakers: Lisa Moutzouris & Duncan Young (Australian Bureau of Statistics) 


Topic: 2016 Census – bigger Australia, smaller Census


Abstract: On 9th August 2016 the Australian Bureau of Statistics (ABS) will be conducting the 17th Australian Census. This Census will count close to 10 million dwellings and approximately 24 million people in Australia on Census night, but will be conducted significantly differently from previous Censuses. The ABS is using behavioural economics, a national address register, behavioural predictions and real-time management information to encourage more than 15 million people to complete the Census online and deliver the Census for over $100m less than under the traditional Census model.


Lisa Moutzouris

Lisa Moutzouris is the Director of the 2016 Census in South Australia. Lisa has a Bachelor of Arts from Flinders University with majors in the social sciences.  Lisa has spent the majority of her working career in the ABS, a career spanning over 25 years.

In that time Lisa has worked in many areas of the ABS including client servicing, household surveys, culture and recreation, and regional statistics. She has a very good understanding of clients’ needs for data and data application. Lisa spent 12 months outposted to a state government department where she gained valuable experience in the use of data to inform policy decisions.


As the Director of the 2016 Census Lisa is leading a team of 30 office based and over 2,000 field staff to ensure every person and every dwelling in SA on Census night is counted.


Duncan Young

As a graduate of computer science and mathematics, Duncan Young has undertaken a range of technology and statistical leadership roles at the Australian Bureau of Statistics and Statistics New Zealand over the last 15 years. Duncan is responsible for delivering the Australian Census of Population and Housing in 2016.


Duncan was recognised as the 2011 Australian Government Young ICT Professional of the Year, completed his Masters of Public Administration through the Australia and New Zealand School of Government in 2012 and received the 2013 Young Public Sector Leader Award.


In addition to his role with Australia, Duncan has provided direct assistance to the Vietnam Census in 2009, the UK’s Beyond 2011 Review, Myanmar’s Census in 2014 and the UN Statistics Division’s Principles and Recommendations for the 2020 round of Censuses.




The AGM will be held on Wednesday 23rd March 2016.




Feel free to forward this meeting notice to colleagues, all welcome.




Paul Sutcliffe


SA Branch

Statistical Society of Australia Inc

Tel. 0438 446 064

[email protected]




Cornish Memorial Lecture Announcement

South Australian Branch of the Statistical Society of Australia


The South Australian Branch of the Statistical Society warmly invite you to attend the 2015 E.A. Cornish Memorial Lecture, to be given by Professor John Carlin. Held every two years, this lecture is the highlight of our speaker program and we look forward to seeing you there.


Date:  WEDNESDAY 18 November 2015


Venue:     Napier 102, University of Adelaide, North Terrace. See the campus map, available at .


Time:   5.15pm        Pre-meeting drinks & nibbles (foyer).

6.05pm   E. A. Cornish Memorial Lecture.

7.30pm        A banquet dinner ($48 plus drinks) will be held in the private function room at Jasmin Indian

Restaurant, 31 Hindmarsh Square, Adelaide.


Please RSVP for dinner and advise of any special dietary requirements to

[email protected]  by Wednesday 11th November 2015.


Speaker:  Professor John Carlin. Director, Clinical Epidemiology and Biostatistics Unit, Murdoch Children’s Research Institute, Melbourne.


Topic: “Statistics and statisticians in real-world research: science or snake-oil?”



Abstract: The past few years have seen increasing discussion in the broader academic community about the so-called “crisis of reproducibility” in science. For example, a recent paper (Nosek et al, Science, 2015) reported on a large-scale project in which 100 original studies in psychology were replicated, with the result that the average effect size reduced by a half in the replication studies. Many writers have suggested that the widespread belief that statistical significance provides a conventionally accepted licence for making scientific claims from noisy data is a major factor in the flakiness of scientific reporting. This has led to some extreme reactions, such as the decision of a psychology journal’s editors in 2014 to place a complete ban on the use of significance testing (and in fact of all formal statistical inference) in research articles submitted for publication. Such extreme reactions may provoke derision from statisticians but it should also prompt us to examine carefully our role in broader scientific discourse. It can be argued that statisticians have not generally paid sufficient attention to these problems, and that by teaching and promoting idealised and readily corruptible technical tools for performing statistical inference we may as a profession have aided and abetted poor scientific practices. I will review these current debates about scientific reproducibility and the role of statistics, and conclude by suggesting that statisticians could pay greater attention to and take greater responsibility for how basic statistical concepts are used in practice.



Biography: John Carlin is Director of the Clinical Epidemiology and Biostatistics Unit within the Murdoch Children’s Research Institute and University of Melbourne Department of Paediatrics at the Royal Children’s Hospital, Melbourne, and has a professorial appointment in the Centre for Epidemiology & Biostatistics, School of Population & Global Health, University of Melbourne. Since completing a PhD in Statistics at Harvard University in the 1980s, John has worked mainly in biostatistics, engaging in collaborative research across a broad range of topics in medicine and public health, while also maintaining some practically oriented statistical research in longitudinal data and multiple imputation for missing data problems. He has co-authored more than 330 scientific publications, and was a founding member of the Steering Committee of the Biostatistics Collaboration of Australia, which delivers a Masters program in biostatistics nationally. He currently leads the Victorian Centre for Biostatistics (ViCBiostat), funded as an NHMRC Centre of Research Excellence, in collaboration with colleagues at Monash University and the University of Melbourne. In an earlier (or parallel?) life, he co-authored an influential textbook, Bayesian Data Analysis (first author Andrew Gelman), which is now in a 3rd edition.



The E.A. Cornish Lecture Series


The Statistical Society of Australia, South Australian Branch, inaugurated a series of public lectures on statistical topics of broad interest in 2001. The lecture series has been named to commemorate Alf Cornish, a leading figure in the early years of the statistical profession in Adelaide.


The lectures are held biennially and presented by eminent statisticians from around the world. Previous presenters of the Cornish Lecture have been Professor Terry Speed on the topic “Gene Expression”, Professor Adrian Baddeley on “Practical analysis of spatial points patterns”, Professor Kerrie Mengersen on “Making Decisions Based on Data”, Denis Trewin “Statistical Critique of the International Panel on Climate Change’s work on Climate Change”, Dr Louise Ryan “Data, data everywhere!”, Peter Diggle on “A Tale of Two Parasites: Model-based Geostatistics and River Blindness in Equatorial Africa”, and Noel Cressie “Statistical Science: A Tale of Two Unknowns”.



Edmond Alfred Cornish (1909 – 1973)


E.A. Cornish graduated from Melbourne University in 1931 with first class honours in Agricultural Biochemistry, Agricultural Engineering and Surveying. While working as an agrostologist (specialist in grasses) at the Waite Research Institute, a centre for agricultural research and development in Adelaide, he became interested in statistical issues arising in agriculture. His interest in patterns of rainfall and their relationship to the yield of natural pastures continued throughout his life.


In 1937 he took a leave of absence at his own expense to study statistics with R.A. Fisher in London. On his return, he was appointed statistician to the Waite Institute. In 1940 he was appointed as Officer-in-Charge of the Biometric Section of the Council for Scientific and Industrial Research (CSIR, now CSIRO) in Melbourne. Under his leadership, the Biometric Section grew, attracting such high calibre scientists as Evan Williams, George McIntyre and Helen Newton Turner. In 1944 the headquarters of the Section was moved to Adelaide and renamed the Mathematical Statistics Section; in 1954 it became the Division of Mathematical Statistics (DMS), with Cornish as its first Chief. Under his leadership DMS grew to 50 staff at his death in 1973.


During the late 1950’s, the University of Adelaide had become aware of the importance of mathematical statistics and appointed Cornish as Foundation Professor of Mathematical Statistics at the University of Adelaide from 1960 until 1965, when his former student Alan James returned from Yale to take over the role.


While his name is perhaps most often heard in connection with the Cornish-Fisher expansion of quantiles of the distribution of a mean in terms of cumulants, his contributions to statistics and the profession were broad and of considerable significance for the development of statistics in Australia.


In addition to his early work on rainfall, he published extensively on experimental designs and analysis of experimental data, particularly in the presence of missing values. His work with Fisher led him to a strong interest in fiducial theory. This led him to develop ground-breaking ideas in multivariate analysis, including the development of a multivariate t -distribution to obtain fiducial distributions of multivariate means.


He was enthusiastic about the use of electronic computers in statistical work, perhaps as a result of his work on climatology, which involved the calculation and modelling of 90585 correlation coefficients. He appreciated early the potential for simulation to answer intractable statistical problems, and promoted the establishment of CSIRO’s Division of Computing Research, whose successor, the Division of Information Technology joined with DMS to form the current Division of Mathematical and Information Sciences.


He was a Fellow of the Australian Institute of Agricultural Science, an Honorary Fellow of the Royal Statistical Society and a Fellow of the Australian Academy of Science. He also served as President of the international Biometric Society and of the Australasian Region.


Alf Cornish laid the foundations for the strong tradition of experimental and theoretical statistics in Adelaide and it is fitting that his name should be associated with a series that will bring eminent statisticians to Adelaide to support the ongoing strength of the statistical profession here.

Upcoming Meetings


The South Australian Branch of the Statistical Society would like to invite you to the October talk of the 2015 program.

Venue: Ingkarni Wardli, Level 7 Conference Room, North Terrace, Adelaide University.

Campus map is available at

***Please note that most entrance doors to Adelaide University buildings close at 6pm so make sure you arrive in time for the talk.


5.30pm – Refreshments.

6.05pm – General Meeting Talk.

7.30 pm – A dinner will be held after the meeting at Lemongrass Thai Bistro, 289 Rundle Street, Adelaide.

Please rsvp for dinner to [email protected] by COB Mon 26th October.


Speaker: Associate Professor David Lynn (SAHMRI) 

Topic: Using networks to investigate the dysregulation of cellular responses in disease


Abstract: Network biology is a rapidly developing area of biomedical research, reflecting the appreciation that complex phenotypes, such as disease susceptibility, are not the result of single gene mutations acting in isolation but rather are due to the perturbation of a gene’s network context. Understanding these molecular networks and the nodes which are key to their topology, structure and signalling is seen as key to understanding complex systems. In this talk, I will give examples of two projects where we are using networks to help us understand how cellular responses are dysregulated in disease. The first is the InnateDB project, a computational platform facilitating network based analyses of the mammalian innate immune system, and the second is the PRIMES project, in which we are investigating how a signalling pathway is re-wired in cancer.


Biography: Associate Professor David Lynn is an EMBL Australia Group Leader in the Infection and Immunity Theme at the South Australian Health and Medical Research (SAHMRI), with a joint faculty appointment as Associate Professor at the School of Medicine, Flinders University. David’s primary research interest is investigating the regulation of the immune system from a genome-wide or systems level perspective. On the wet-lab side, the group employs in vitro and in vivo experimental models coupled with systems biology approaches to investigate the regulation of innate &, more recently, adaptive immunity. On the bioinformatics side, the group leads the development of, an internationally recognised systems biology platform for the computational analysis of innate immunity networks/pathways. Recently, David has also expanded his interest in network biology into the cancer signalling area, and lead the computational biology aspects of €12 million European Commission funded project, investigating how to model and subsequently therapeutically target networks in cancer. To facilitate this work the group has developed several novel pieces of software for network and pathway visualisation and analysis.


Feel free to forward this meeting notice to colleagues, all welcome.


*** Save the Date ***
The 2015 EA Cornish Memorial Lecture will be given by John Carlin on Wednesday November 18.  The title of John’s talk is “Statistics and statisticians in real-world research: science or snake-oil?”.  Further details to come.


The SA Branch of the Statistical Society is promoting the upcoming SAS User Group meeting (see Below) and the Young Statisticians Career Event.

Young Statisticians’ Careers Event

 Tuesday October 13

Food, drinks and networking from 5:15pm to 6:00pm

Panel Q & A with employers from 6:00pm to 7:30pm

University of Adelaide, North Terrace Campus

Lower Napier LG28


The South Australian Branch of SSAI would like to invite all students working in the field of statistics for an informal gathering to meet other students, potential future employers, and members of the society for a friendly chat over FREE FOOD AND DRINK*, followed by a panel Q & A with employers to learn where your degree could take you.

Recent statistics graduates and post-graduate students doing statistics in other fields such as econometrics, social sciences, and health sciences are welcome.

RSVP: [email protected] by Friday October 9


*Gluten free and vegetarian options will be available – please specify any dietary requirements in your RSVP. Non-alcoholic beverages will also be provided.

This event is sponsored by SAS Users South Australia (SUSA), and follows from their event, which begins at 3:30pm. To register for their event, go to





The South Australian Branch of the Statistical Society would like to invite you to the next talk of its 2015 program on Wednesday 2nd September.
Venue: Engineering North Building, Room N132, North Terrace, Adelaide University.
Campus map is available at
***Please note that most entrance doors to Adelaide University buildings close at 6pm so make sure you arrive in time for the talk.
5:30pm – Refreshments.
6:05pm – General Meeting Talk.
7:30pm – A dinner will be held after the meeting at Cafe Michael 2, 204 Rundle Street, Adelaide.
Please RSVP for dinner to [email protected] by Monday 31st August.
Speaker: Phil Bell, Australian Bureau of Statistics
Topic: An interactive approach to the multivariate allocation of resources
Resource allocation considers how to divide limited resources between tasks so as to achieve the best outcomes.  In a stratified sample survey the decision to be made is how many units to select from different groups (strata) in the population.  The outcomes to be assessed in designing the survey are the relative standard errors (RSEs) that are predicted to result for a set of key survey estimates.

A typical approach to this problem is to optimise a single cost function subject to multiple RSE constraints. There is no closed form solution to this problem.  More importantly, the required cost may not fit within available funding, leading to a cycle of revisions of the constraints in consultation with survey users.  The situation can be complicated further where there are multiple costs to be considered.

I explore an alternative approach that acknowledges this interactive nature of the allocation problem. The approach starts with a compromise allocation that the survey designer can modify towards meeting the requirements of the survey. The calculations required are simple enough to be performed in a spreadsheet.  My implementation uses an Excel spreadsheet – this provides immediate graphical feedback of the impact of varying the importance of any particular estimate. The resource allocation given at any iteration is optimal in the sense that there is no lower-cost allocation that predicts the same set of RSEs (or any set that is uniformly better).

In this talk I will discuss this interactive approach to resource allocation, and demonstrate it using the sample allocation spreadsheet..
Biography: Phil Bell said good-bye to full-time study in 1983 when he graduated from Sydney University with a first class honours degree in Mathematical Statistics. He has spent his career working in the Methodology Division of the Australian Bureau of Statistics, and this has given him the opportunity to develop statistical methods across a range of applications in survey statistics and data analysis. His contributions were recognised in January 2015 by an Australian Statistician’s Award for Leadership through innovation.  .


*** Save the Date ***
The 2015 EA Cornish Memorial Lecture will be given by John Carlin on Wednesday November 18.  The title of John’s talk is “Statistics and statisticians in real-world research: science or snake-oil?”.  Further details to come.



The South Australian Branch of the Statistical Society would like to invite you to the July talk of the 2015 program.

Venue: Napier Building, Room LG24, North Terrace, Adelaide University.

Campus map is available at


***Please note that most entrance doors to Adelaide University buildings close at 6pm so make sure you arrive in time for the talk.



5.30pm – Refreshments.

6.05pm – General Meeting Talk.

7.30 pm – A dinner will be held after the meeting at Jasmin Indian Restaurant, 31 Hindmarsh Square, Adelaide.

Please rsvp for dinner to [email protected] by COB Mon 27th July.


Speakers: John Gray and Craig Hansen (SAHMRI) 


Topic: Producing South Australian small area estimated resident population (ERP) counts for 5 year age/sex and Indigenous status for intercensal years


Abstract: Estimating population counts for the Aboriginal and Torres Strait Islander population has a number of challenges due to the uncertain quality of attribution of Indigenous status in baseline demographic data sets such as births and deaths, inconsistency in Indigenous identification over time and apparent under counts of Aboriginal and Torres Strait Islander people in census data.

In addition, where estimated resident population counts for Aboriginal and Torres Strait Islander people are produced by the Australian Bureau of Statistics between intercensal years, generally, these are only available by age and sex for large geographic units such as “all of state” and remoteness level or at small area level for the total population.

The Landscape project seeks to produce contemporary population level health and health related data at the small area level including population rates for selected measures. This presentation will describe the method used for creating meaningful geographic units of analysis (Landscapes) within South Australia and then generating estimated resident population counts by Indigenous status, 5 year age groups and sex at this small area level for the years 2001 to 2013 inclusive.

Some issues in the application and validation of these estimates will be identified for discussion.


Biography – John Gray: John has a BSc. (Hons) in Psychology and a post-graduate Bachelor’s degree in Social Administration.  His earlier career included:

service delivery in adolescent residential care, family day care and education social work

further education delivery, program management and curriculum development in the community services sector

staff development and policy advice in health

Since 2005, John has been a population health analyst and data manager, first for SA Health and since 2013 at SAHMRI in the Wardliparingga Aboriginal Research Unit where he is Manager, Data and Analysis Support and analyst on the Landscape project – a small area study of health status, risk factors and social determinants of health for Aboriginal South Australians.


Biography – Craig Hansen: Craig Hansen gained his doctorate in epidemiology at the University of the Sunshine Coast and has been using SAS for 15 years within the health research setting. In that time, has worked at:

  • the University of Queensland School of Medicine (Australia) – working in cardiovascular research
  • the United States Environmental Protection Agency (USEPA) – working in air pollution research
  • the Centres for Disease Control and Prevention (CDC, USA) – working in birth defects research
  • Kaiser Permanente Centre for Health Research (USA) – working on health studies using electronic medical records.

Craig recently accepted a position at the South Australian Health and Medical Research Institute as Senior Epidemiologist within the Wardliparingga Aboriginal Health Unit with particular responsibility for management and analysis of data for the Aboriginal Diabetes Study (PROPHECY), a cohort study of Type II Diabetes amongst Aboriginal South Australians.


Feel free to forward this meeting notice to colleagues, all welcome.



Paul Sutcliffe


SA Branch

Statistical Society of Australia Inc

Tel. 0438 446 064

[email protected]



Speaker: Dan Navarro (University of Adelaide) 


Topic: What can Bayesian statistics tell us about human cognition?


Abstract: The Bayesian approach to statistical inference treats the problem of data analysis as a form of idealised reasoning, in contrast to the orthodox view that operationalises probability in terms of long run frequency. To a statistician, Bayesian inference is primarily a data analysis tool. In cognitive science, however, Bayesian methods are also used as tools to build theoretical models of how humans learn and reason. In this talk I discuss several applications of Bayesian inference to human cognition, and illustrate how idealised Bayesian models are used to help make sense of the implicit assumptions that underpin human reasoning.


Biography: Dan is an Associate Professor and ARC Future Fellow at the University of Adelaide, and a director of the Computational Cognitive Science lab there. His research focuses on probabilistic and information theoretic models of human cognition, with particular emphasis on conceptual learning and decision making. Current topics of research in the lab include pragmatics and human reasoning, category learning, sequential decision making and the explore-exploit dilemma, language evolution, human information search and hypothesis testing, and many others.


Feel free to forward this meeting notice to colleagues, all welcome.


Past Meetings



Venue: Ingkarni Wardli Building, Room B19, North Terrace, Adelaide University.

Campus map is available at


***Please note that most entrance doors to Adelaide University buildings close at 6pm so make sure you arrive in time for the talk.



5.30pm – Refreshments.

6.05pm – General Meeting Talk.

7.30 pm – A dinner will be held after the meeting at Lemongrass Thai Bistro, 289 Rundle Street, Adelaide.

Please rsvp for dinner to [email protected] by 24th May.

Speaker: Tessa Longstaff (University of Adelaide) 

Topic: Analysis of Longitudinal Data when the Length of Follow Up is Informative


Abstract: Most commonly used statistical methods assume that the data consist of independent observations. When considering data for which this assumption is not valid, other methods of analysis are needed. One example of non-independent data occurs in longitudinal studies, where the variables are repeatedly measured over time on each subject. Additional complexity can arise if the data has informative cluster size. That is, if the number of observations per subject is related to the outcome.  Although many methods for dealing with dependent data have been developed, most are unsuitable for data with informative cluster size. In this talk I will be considering the problem of informative cluster size in longitudinal data in the context of real and simulated data.


Biography: Tessa is in her second year of a Masters of Philosophy at the University of Adelaide, working in Statistics. She completed a Bachelor of Mathematical Sciences at the University of Adelaide in 2013. Her main interest is in clustered data analysis.


Feel free to forward this meeting notice to colleagues, all welcome.



Paul Sutcliffe


SA Branch

Statistical Society of Australia Inc

Tel. 08 8298 8179

[email protected]



The South Australian Branch of the Statistical Society would like to invite you to the April talk of the 2015 program.

Venue: Engineering North Building, Room 132, North Terrace, Adelaide University.

Speakers: Professor Ben Mol (University of Adelaide) 

Topic: A day without randomisation is a day without progress

Abstract: When it comes to receiving medical treatment, the majority of us would make a fundamental assumption: sound evidence exists to suggest the proposed course of action will improve our condition. Here in South Australia, however, that assumption is too often wrong. Just as everywhere in the world, around 50% of our $5 billion health care budget facilitates the delivery of medical interventions that we cannot say with empirical certainty will make us healthier.

But it needn’t be this way. In many countries, including Australia, ongoing evaluation of the effectiveness of interventions is now prioritised within health systems themselves, rather than being dependent on externally funded research. With a particular focus on the area of reproductive health, this important presentation will show how this approach can lead to the creation of a globally self-learning medical system, with substantial benefits for patients, communities and governments alike. I will specifically focus on the role of medical statistics in this matter.

Professor Mol has worked as a gynaecologist and senior researcher at the Utrecht Medical Centre (1997-2003), Maxima Medisch Centre (2003-2007) and the Academic Medical Centre (2007-2014) in the Netherlands. He moved to Adelaide in 2014 to become Professor of Obstetrics and Gynaecology at the University of Adelaide. His research is focused on the organisation of multi-centric evaluative research in obstetrics, gynaecology and fertility. Professor Mol has co-authored > 700 peer reviewed publications, and his H-index is 49.  His professional adage is “A day without randomisation, is a day without progress”.



Annual General Meeting Announcement of the South Australian Branch of the Statistical Society of Australia  WEDNESDAY 25 MARCH 2015

VENUE: University of Adelaide, North Terrace  Meeting room on Level 7 in Ingkarni Wardli Building


Election of Returning Officer, Branch officers, Council and Public Officer

Please note only financial members of the society will be eligible to vote in person, or by proxy.

While a number of council members are willing to stand for office again, this is a first call for nominations for the positions of:

President, Vice-President, Secretary and Treasurer and, Councillor positions assist in the following tasks related to local activities and the relationship with the central SSAI:

Newsletter correspondent, Young Stats organiser(s), Website coordinator, Speaker program manager

Nominations are called for all positions and may be communicated to the Secretary Paul Sutcliffe ([email protected] or 82988179) before the meeting or may be made from the floor at the meeting.


  1. Apologies
  2. Minutes of the 2014 Annual General Meeting (attached)
  3. Annual Report
  4. Election of Returning Officer, Branch officers, Council and Public Officer
  5. Treasurers Report
  6. Election of Auditor
  7. Awards
  8. Any other business
  9. Speaker: Brock Hermans (University of Adelaide) “Inference for epidemics on networks”

Speaker:  Brock Hermans – University of Adelaide.

Title:  Inference for epidemics on networks

Abstract:    The motivating question for many epidemiologists is \what is the infection rate”? That is, what is the rate at which people become infected by a particular disease? As we form more realistic epidemiological models estimating the infection rate becomes more complex.

Estimating the infection rate is straight forward for the usual susceptible-infective-recovered (SIR) model. However this model assumes that an individual in a population is equally likely to have contact with everyone else in the population, and vice versa. The assumption of equal contact is unrealistic in most real-world settings, and so we turn to a different model.

In this talk I explain a method for generalising the SIR model that uses network theory. Networks allow us to generalise the idea of contact between people by looking at nodes (people) and edges between them (representing a connection). Here, a connection between nodes represents a possible path of infection. I will then look at a simple way of estimating the infection rate based on the idea of using the usual SIR framework as an estimate of the infection rate. We then adjust this estimated value based on some properties of our network.

Biography: Brock is in his second year of a Masters of Philosophy at the University of Adelaide, working in Statistics and Applied Mathematics. He completed a Bachelor of Mathematical Sciences at the University of Adelaide in 2013. His main interest is in solving stochastic systems using statistical methods.



The South Australian Branch of the Statistical Society in conjunction with the Young Statisticians Conference 2015 would like to invite you to a public lecture given by Sheila Bird. All welcome.

Venue: Horace Lamb Lecture Theatre, North Terrace, Adelaide University

Speaker: Professor Sheila Bird (MRC Biostatistics Unit, Cambridge) 

Topic: Biostatistician behind bars.

Abstract: The UK’s Medical Research Council Biostatistics Unit celebrated its centenary in 2014. Over its history, the Unit has tackled epidemics from tuberculosis to HIV, cigarette smoking to heroin injection, and their related causes of death.  Heroin injectors are, of course, frequently incarcerated for acquisitive crimes and vulnerable to blood-borne viruses.

I shall describe how a quarter century of surveillance designs (with associated biological sample), record-linkage studies, and bespoke “questionnaries”  – a phrase coined from Hill and Doll – have improved prisoners’ access to harm reduction (eg Hepatitis B immunization) , contributed to changed policy in prisons (eg put an end to random mandatory drugs testing), got us barred from doing studies in prisons, but led us to quantify a 7 times higher risk of overdose death soon after prison-release, and eventually enabled three musketeers to mount the pilot N-ALIVE Trial in England, which tests whether those randomized to receive naloxone-on-release have 30% fewer opioid-related deaths in the 4-weeks post-release than controls (prior estimate: 1 in 200).

Even before the N-ALIVE Trial’s first randomization, however, Scotland became the first country to introduce take-home-naloxone as a funded public health policy. Wales followed suit in May 2011.  I shall describe the trials and tribulations of Scotland’s closely-monitored evaluation of its 2011-15 take-home naloxone policy, which is complicated because Scotland’s policy was introduced against a still-rising trajectory of age-related opioid-deaths.

I shall also describe the impact of the Scottish results on the pilot N-ALIVE Trial and ask: is it impact for a randomized trial to be overtaken by the policy it seeks to inform?

As a programme leader at MRC Biostatistics Unit, Sheila Bird applies statistical science, including record-linkage, at the interface of public health and other jurisdictions. Her focus is on drugs-related deaths, their projection, and how to reduce them; the sentencing and health hazards that drug-treatment clients face in the short and longer-term; and better surveillance for, and quantitative understanding of, alcohol as co-morbidity (for HCV progression) or contributor to cause-specific mortality in Scotland.In the context of court-based randomization, record-linkage is an affordable, unbiased, under-utilized means for assessing the cost-effectiveness of sentencing. Sheila holds a visiting professorship at Strathclyde University’s Department of Mathematics and Statistics, She was awarded the RSS’s Guy (bronze), Bradford Hill and Chambers Medals in 1989, 2000 and 2010, made OBE in 2011, and elected a Fellow of the Royal Society of Edinburgh in 2012. Sheila was a Medicines Commissioner and first statistician on the NICE Appraisal Committees, and chaired RSS’s Working Party on Performance Monitoring in the Public Services.


End of Year BBQ

Wednesday December 10th at 5:30pm

University of Adelaide, North Terrace Campus

Level 5, Ingkarni Wardli

The South Australian Branch Council would like to invite all members and guests to an end of year BBQ. Come and join your statistical colleagues for a friendly chat and catch up before the year is over.

The society will provide meat, salads, nibbles, wine and soft drinks – BYO other alcohol/drinks. All you have to do is rsvp and turn up.

Everyone is welcome.




Venue: Horace Lamb Lecture Theatre, University of Adelaide, North Terrace.

Speaker: Professor Terry Speed, Head of Bioinformatics, Walter and Eliza Hall Institute of Medical Research will give a public lecture as part of his AMSI- SSAI lecture tour – all are welcome.

Title: A New Frontier: understanding epigenetics through mathematics

Abstract: Scientists have now mapped the human genome – the next frontier is understanding human epigenomes; the ‘instructions’ which tell the DNA whether to make skin cells or blood cells or other body parts. Apart from a few exceptions, the DNA sequence of an organism is the same whatever cell is considered. So why are the blood, nerve, skin and muscle cells so different and what mechanism is employed to create this difference? The answer lies in epigenetics. If we compare the genome sequence to text, the epigenome is the punctuation and shows how the DNA should be read. Advances in DNA sequencing in the last five years have allowed large amounts of DNA sequence data to be compiled. For every single reference human genome, there will be literally hundreds of reference epigenomes, and their analysis could occupy biologists, bioinformaticians and biostatisticians for some time to come.



Venue: CSIRO Reception (F3 on grid) Soil & Water Environs Centre (Building C1) – Large Meeting # 1, Enter via Gate 4, Waite Rd, Urrbrae.

Speakers: Dr Brett A. Bryan (CSIRO)   

Topic: Influence of policy on land use and ecosystem services: sustainability under global change

Abstract: The introduction of climate (e.g. a carbon market) and energy policy may provide significant opportunities for the widespread adoption of new land use and management options for greenhouse gas (GHG) mitigation in agricultural landscapes. New land uses may include a range of bioenergy, carbon plantings, biodiverse plantings, and alternative crop and livestock management uses. However, these changes may generate collateral impacts (positive and negative, direct and indirect) for regional development, energy security, food production, land and water resources, biodiversity conservation, and other ecosystem services. There is a real need for the ex-ante evaluation of the impact of alternative policy options under a range of biophysical and economic scenarios on ecosystem services in agricultural landscapes. Consideration of the uncertainty underpinning the range of input data and models and the impacts of future scenarios is essential to understand the robustness of decisions and outcomes. This information is required to inform decision-making for the transition to a low carbon economy.

I will present the theoretical context underpinning the connection between policy, land use, and ecosystem services, and its place in the global science-policy interface. I will take the audience on a tour of the evolution of a decade of work on the integrated modeling of land use and ecosystem services in Australia—from catchment, regional, to national scale. I will then finish with some outcomes of land use scenarios from the Land Use Trade-offs (LUTO) model—part of the Australian National Outlook. Global economic and climatic trends are estimated using an integrated assessment model. LUTO itself integrates a wide range of biophysical and economic data and models in a spatio-temporal model of potential land use change in Australia’s intensive agricultural zone out to 2050. LUTO itself is a high resolution partial equilibrium model of the agricultural sector. It allocates land use including existing agriculture but also new land uses such as environmental plantings, carbon plantings, woody perennials and cereal crops for bioenergy and biofuels. The implications of policy options and changes in external drivers on land use and the impact on across a range of ecosystem services is assessed including food and fibre production, carbon, water, energy, and biodiversity. We analyse four global scenarios—one low emissions, two medium emissions, and one high emissions pathway. We consider climate change projections from three GCMs. Also, assessed are three biodiversity payment policies, three productivity scenarios. Selected outcomes from these national land use outlooks will be presented, including a deep-dive into the trade-offs between carbon sequestration and biodiversity services. Model outcomes inform policymakers of the costs, benefits and trade-offs associated with specific policy directions under global change and other elements of deep uncertainty..

Dr Brett A. Bryan is a Principal Research Scientist and project leader in CSIRO—Australia’s pre-eminent public scientific research institution. His research is focused on creating cost-effective policy for the sustainability of social-ecological systems. As a geographer, Dr Bryan has research interests in the application and development of computational tools and analytical methods in a diverse array of social and environmental contexts. He has expertise and over 20 years experience in integrated environmental-economic modelling over space and time. He has domain expertise in integrated modelling of land use and ecosystem services, and has worked in both terrestrial and marine environments. Dr Bryan’s research interests are at the human/environment interface combining aspects of land use and management; agriculture and food security; water resources management; climate change impact assessment, mitigation, and adaptation; biodiversity conservation; energy and lice-cycle analysis; and economic and policy analysis. He has conducted research in China, India, Indonesia, the United States, and many parts of Australia. He has published nearly 70 papers in the international scientific literature. Currently, he is leading the development of the Land Use trade-Offs (LUTO) model, assessing future scenarios for the sustainability of Australia’s agricultural regions.



Venue: Ingkarni Wardli Building, Room 5.57, North Terrace, Adelaide University.

Speakers: David Price & Ben Rohrlach (University of Australia)  

Topic: Optimal Experimental Design for Group Dose-response Challenge Experiments

Abstract: An important component of scientific research is the ‘experiment’. Effective design of these experiments is important and, accordingly, has received significant attention under the title ‘optimal experimental design’. However, until recently, little work has been done on optimal experimental design for experiments where the underlying process can be modelled by a Markov chain. In this talk, I will briefly discuss some background and methods for optimal experimental design, including a new method we have developed for finding optimal designs. I will discuss some results from these methods when applied to group dose-response challenge experiments for the bacteria Campylobacter jejuni in chickens.

David Price completed his Bachelor of Mathematics with Honours in Statistics at the University of Adelaide in 2011. He is currently in the third year of his PhD at University of Adelaide, looking into optimal experimental design for Markov Chain models.

Topic: Data Driven Model Selection with Approximate Bayesian Analysis

Abstract: Approximate Bayesian Computation (ABC) is a likelihood-free method for obtaining samples from the posterior distribution of some parameter set, and is useful when the likelihood function is difficult, or impossible, to obtain. The basic concept behind ABC is that we compare a large number of simulated data sets to our observed data set and retain some as our sample form the posterior distribution. In order to perform the comparison, summary statistics are generally used, and these summary statistics are rarely sufficient. In this talk I aim to introduce ABC, and motivate its use in analyses. Finally I will describe a method for using this same “observed data vs simulated” idea from ABC to perform data driven model selection on our data

Ben Rohrlach is PhD Candidate, Mathematical Sciences, B.Ma.Sc, M.Phil(Stat) at the University of Adelaide.



Venue: Ingkarni Wardli Building, Basement Room 18, North Terrace, Adelaide University.

Speaker: Professor Andy Koronios (University of South Australia) 

Topic: Upskilling in Data Science in South Australia, for professionals and students.

Abstract: It is has been reported that the age of Big Data will change our world and generate thousands of jobs and the career of the data scientist may already be one of the most sought-after people in the ICT sector. What are the skills of becoming a data scientist? And how can they be acquired to give practitioners the edge in leveraging the benefits of such great possibilities in this field? UniSA was one of the first higher education institutions to introduce a suite of programs in Data Science. The structure and attributes of these programs will be discussed during this presentation.

Note: This talk is a repeat of the talk given at the recent IAPA meeting in July which clashed with the Statistical Societies July talk.

Professor Andy Koronios is the Head of the School of Information Technology & Mathematical Sciences, at the University of South Australia. Andy holds academic qualifications in Electrical Engineering, Computing and education and a PhD from the University of Queensland.

Andy has extensive experience in both commercial and academic environments and has research interests in information management & governance, e-commerce and the management and strategic exploitation of information as well as the changing role of the IT Manager. Andy has established two University Research Labs and a funded Research Centre and is currently the Director of the Strategic Information Management Research Lab in the Advanced Computer Research Centre. He has worked both as a consultant as well as a professional speaker on IT issues in Australia and South East Asia and has over twenty years’ experience in the academic environment. He is a Fellow of the ACS, and a Founding Fellow of the International Institute of Engineering Asset Management Editor-In-Chief of the International Journal of Information Quality and Associate Editor of the ACM Journal for Information Quality.



Venue: Ingkarni Wardli Building, Basement Room 18, North Terrace, Adelaide University.

Speaker: Professor Lyle Palmer (Joanna Briggs Institute, University of Adelaide)  

Topic: Analytic issues related to large-scale longitudinal cohort studies.

Abstract: Progress towards rapid and effective translation of new information into clinical and public health applications will ultimately depend upon the availability of large and well-characterized population-based cohorts studies, underpinned by total population, longitudinal data and family record linkage so that all those with and without disease, their risk and protective factors, including both genetic and environmental contributions, can be studied in an unbiased way throughout the whole life span. Such underpinning linkages and resources already exist in Ontario. We are in the process of developing a national and international research platform that will build on unique Ontario population health data collected and managed over the last three decades. The Ontario Health Study (OHS) is a population-based cohort study that serves as an integrated platform for a wide range of studies that explore how lifestyle behaviours, physiological measures, genetics, and community-level factors contribute to the development of cancer, heart disease, diabetes, asthma, and other chronic diseases. The OHS ( is the largest volunteer cohort study ever conducted in Canada, and seeks to collect extensive data from every consenting Ontarian 18 years of age and over – a sampling frame of ~9.5 million people across the province. The OHS will follow participants for their entire lifespan through both active and passive (data linkage) follow-up. An online questionnaire is used to collect detailed information at baseline from participants regarding their socio-demographic background; personal and family medical histories; and health behaviours. Additional measures collected during active follow-up will include assessments of diet, various aspects of mental health, environmental exposures, and residential and occupational histories. Subsets of volunteer participants are also having physical assessments and donating a biospecimen. A critical current challenge is to develop the biostatistical and bioinformatics tools and resources necessary to fully utilize and grow the OHS into an internationally competitive research resource. The challenges posed by high-dimensional data of this scale and depth and by the causal pathway paradigm for statistical analysis are profound.

Key words: Cohort study; survival analysis; bias; sampling; longitudinal analysis.

Professor Lyle Palmer has recently relocated to Adelaide from Toronto to take up a new opportunity as Executive Director of the Joanna Briggs Institute at the University of Adelaide. Before moving to Adelaide, Professor Palmer was a Senior Principal Investigator and Program Director at the Ontario Institute for Cancer Research, and a Professor of Biostatistics, Epidemiology and Obstetrics & Gynaecology at the University of Toronto. Together with many partner organizations across Ontario, Professor Palmer led a large-scale expansion of the provincial capacity in the area of translational epidemiology. From 2010 to 2014, he was the founding Executive Scientific Director of the Ontario Health Study (, the largest population-based cohort study (n=230,000) ever undertaken in Canada. Professor Palmer is also a key member of the team currently conducting the Ontario Birth Cohort, which is designed to be one of the largest and best characterized birth cohorts in the world. Prior to moving to Canada, Professor Palmer was the foundation Winthrop Chair in Genetic Epidemiology and the founding Director of the Centre for Genetic Epidemiology & Biostatistics at the University of Western Australia, where he was also a Professor in the Schools of Medicine & Pharmacology and Population Health. Until 2003, he was an Assistant Professor of Medicine at Harvard Medical School and the Director of Statistical Genomics at the Channing Laboratory, Boston. His background includes training in epidemiology, human genetics and biostatistics. He has a particular interest in the areas of life-course genetic epidemiology and the developmental origins of health and disease (DoHAD).

Professor Palmer has been recognized for his leadership role in biomedical research by numerous awards, including Fulbright and Churchill Fellowships. Over the last 10 years, he has chaired and/or given invited symposia at over 30 international scientific meetings, has delivered over 200 invited lectures, has produced over 200 publications, and has co-edited a commercially successful encyclopedia of genetic epidemiology that has become a standard reference work. He is in high demand internationally as a speaker and teacher.



Venue: Ingkarni Wardli Building, Basement Room 18, North Terrace, Adelaide University

Speaker: Dr Lisa Yelland (University of Adelaide) 

Topic: Analysis Issues in Perinatal Trials with Multiple Births.

Abstract: Perinatal trials including infants from both single and multiple births present unique statistical challenges. Cluster randomisation is common in this setting to ensure that all infants from the same birth receive the same treatment. While methods for analysing clustered data are widely available, perinatal trials are somewhat different to the usual clustered data settings due to the small cluster sizes. Many clusters consist of only a single infant, which results in a mixture of independent and clustered data. Conflicting recommendations have been made regarding if and how clustering due to multiple births should be taken into account in the analysis of perinatal trials, particularly when the multiple birth rate is low. The potential for informative cluster size to occur in perinatal trials, where the outcome is related to the size of the cluster, has also recently been recognised and this has implications for the choice of analysis method. In this presentation, I will discuss the different analysis approaches that may be used in perinatal trials including infants from both single and multiple births, and will explore some of the factors that influence whether clustering due to multiple births should be taken into account in the analysis.

Lisa Yelland is an NHMRC Postdoctoral Research Fellow at the Women’s & Children’s Health Research Institute, and in the School of Population Health at The University of Adelaide. She is a biostatistician specializing in maternal and infant health, and a chief investigator on three large perinatal trials. Her research broadly aims to improve the statistical quality of perinatal trials and she is currently investigating methods for designing and analysing such trials when clustering is present due to the inclusion of multiple births.



Venue: Ingkarni Wardli Building, Basement Room 18, North Terrace, Adelaide University.

Speaker: Prof Jerzy Filar (Flinders University) 

Topic: The power and limitations of mathematical models and Plato’s Cave Parable.

Abstract: The rationale for referring to Plato’s Cave Parable is that it describes in a visual and an emotive way what is arguably the essence of the challenge facing most of  the modern era researchers involved in the mathematical modelling of complex phenomena; especially life support systems.  The challenge is that of creating a model whose outputs – Plato’s shadows of images – correspond very closely (under a wide spectrum of inputs) to the measurements of the outputs of the real phenomenon being studied.  For instance, a sound model of the spread of an epidemic in a population should be able to estimate the sizes of the different cohorts affected by the disease, at various stages of the epidemic.  And yet, the mathematical modelling cognoscenti will be conscious of the fact that even a best model of an epidemic is essentially distinct from the epidemic itself.  It is more like a wooden figure of an animal in Plato’s parable than the animal itself.

The societal reliance on mathematical models to support planning, technological innovation, engineering design, and business and development practices is greater than ever before in the history of civilisation.  Furthermore, as availability of high speed computing increases, this trend can only continue.  Therefore, the question addressed in this talk is not whether mathematical modelling is valuable or desirable – that is taken as self-evident – but rather: What are key pitfalls to guard against  when mathematical models of complex phenomena are developed, implemented and used?

**Note this is a repeat of the lecture Jerzy gave at MODSIM2013.

Jerzy Filar is a broadly trained applied mathematician with research interests spanning a wide spectrum of both theoretical and applied topics in Operations Research, Optimisation, Game Theory, Applied Probability and Environmental Modelling. Professor Filar co-authored three research level books. He also authored or co-authored approximately 100 refereed research papers. He is editor-in-chief of Environmental Modelling and Assessment and serves on editorial boards of Operations Research, JMAA and a number of other journals. Professor Filar is a Fellow of the Australian Mathematical Society. He has supervised 19 PhD students who are working at various universities, industries and research institutions across the world.



Venue: Ingkarni Wardli Building, Basement Room 20, North Terrace, Adelaide University.

Speaker: John S. Preisser, Department of Biostatistics, University of North Carolina, U.S.A.  

Topic: Logistic Regression for Dichotomized Counts.


Sometimes there is interest in a dichotomized outcome indicating whether a count variable is positive or zero. For example, oral health researchers are often interested in modeling risk factors, or assessing interventions, for the prevalence of dental caries defined as the proportion of individuals in a population with any dental caries. Under this scenario, the application of ordinary logistic regression may result in efficiency loss, which is quantifiable under an assumed model for the counts. In such situations, a two-part shared-parameter hurdle model is investigated for more efficient estimation of regression parameters relating to overall effects of covariates on the dichotomous outcome, while utilizing count data with many zeroes. Additionally, simulations are used to compare power and Type I error. The methods are applied to data from a randomized clinical trial of three toothpaste formulations to prevent incident dental caries in a large population of Scottish schoolchildren.

John Preisser is Research Professor in the Department of Biostatics, School of Public Health, University of North Carolina at Chapel Hill. His research interests focus on regression analysis for correlated data, in particular, he is interested in the analysis of clustered and longitudinal data, methods for count data including zero-inflated Poisson and negative binomial regression models, statistical methods for missing data, and issues pertaining to the design and analysis of cluster intervention trials.

Annual General Meeting Announcement of the South Australian Branch of the Statistical Society of Australia

Date:           WEDNESDAY 5 MARCH 2014

Venue:        University of Adelaide, North Terrace

Room:         Ingkarni Wardli Building, Basement Room 17


Speaker:  A/Prof Richard Woodman – Flinders University.

Title:  The use of structural equation modelling to examine adiposity and pre-clinical markers of atherosclerosis in young adults

Abstract:    Although structural equation modelling (SEM) is commonly used in psychology and the social sciences its use in areas of epidemiology including that of cardiovascular disease (CVD) is not yet widespread. Instead, most researchers still rely on single regression models to assess the association between exposure and disease outcome. SEM provides a method of simultaneously assessing the complex pattern of inter-relationships that often exist between predictors and outcomes resulting in both direct and indirect effects. Such an approach potentially increases statistical power, affords the testing of causal relationships, and the construct of latent variables based upon multiple measurements which allows the capturing of measurement error. This study explores the use of SEM in estimating the direct effects of adiposity and BMI on intima-media thickening (a surrogate measure of CVD) as well as their indirect effects, via traditional CVD risk factors and novel markers including endothelial function and arterial stiffness. The study uses cross-sectional data from a cohort of young adults in Western Australia.

Biography: A/Prof Richard Woodman is the director of the Flinders Centre for Epidemiology and Biostatistics in the Faculty of Health Sciences at Flinders University. He obtained a PhD in Medicine at UWA studying the effects of different fish oils on CV risk factors in Type 2 diabetes. Much of his early postdoctoral work focused on the reproducibility and utility of novel measures of cardiovascular disease including endothelial function and arterial stiffness. He obtained a Master’s degree in biostatistics from Sydney University and began his career in biostatistics in 2004 at Curtin University. He has a strong interest in sports, particularly cycling and running, previously worked for several years in the fitness industry and holds a Master’s degree in Sports Science from Sheffield University in the UK. As a POM he is also brave enough to currently admit to being a fair weather supporter of the English cricket team and a lifelong “Gooner” i.e fan of the Gunners, (more commonly known as Arsenal FC).


Venue: Ingkarni Wardli Building, Basement Room 20, North Terrace, Adelaide University.

Speaker: Inge Koch, University of Adelaide.  

Topic: Quadratic Forms in Statistics: Evaluating Contributions of Individual Variables.


Quadratic forms capture multivariate information in a single number, making them useful, for example, in hypothesis testing. Other quadratic forms that are commonly used in statistics include the Mahalanobis distance and Fisher’s discriminant function. If the number of variables of the multivariate vector or data is large, or if the statistic obtained from the quadratic form is large, it will be informative to partition the quadratic form into contributions of individual variables.

In this talk I argue that meaningful partitions can be formed, although the precise partition will depend on the criteria used to select it. We consider intuitively reasonable criteria and determine the resulting partition. This partition is based on a transformation of the random vector(s) or data that maximises the correlations between individual variables and the new transformed variables under appropriate constraints. I present properties of the partition including some optimality results. The contributions of individual variables to a quadratic form are less clear-cut when variables are collinear, and forming new variables through rotation can lead to greater transparency. I show how these transformations work in practice in a partitioning of Hotelling’s one- and two-sample T-square statistics, Mahalanobis distance and in discriminant analysis.

This talk is based on joint research with Paul Garthwaite, The Open University, UK. 

Inge Koch studied pure mathematics in Germany and the UK, worked as an applied mathematician in the UK and the CSIRO in Canberra, obtained a PhD in statistics from the ANU, and has since enjoyed life as a statistical mathematician.  After her PhD she taught at the University of Newcastle and the University of New South Wales, and joined the University of Adelaide in 2009 as Associate Professor.  She is fascinated by the mathematical and practical challenges posed by high-dimensional data where classical statistical methods may no longer work and has recently completed the book `Analysis of multivariate and high-dimensional data’.



Venue:  Napier G03, Napier Building, North Terrace, Adelaide University.

Speaker: Patrick Lim, Senior Research Officer, National Centre for Vocational Education Research. 

Topic: The impact of schools on young people’s transition to university.


Patrick is the co-author of a published report on “The impact of schools on young people’s transition to university”. This report uses the Longitudinal Surveys of Australian Youth (LSAY) to look at the impact of schools on a student’s tertiary entrance rank (TER) and the probability of them going to university (controlling for TER).

It shows that the characteristics of schools do matter: although young people’s individual characteristics are the main drivers of success, school attributes are responsible for almost 20% of the variation in TER scores. The three most important school attributes for TER include sector (Catholic and independent/ government); gender mix (single sex/co-educational) and the extent to which a school is ‘academic’.

The socioeconomic status of schools didn’t emerge as a significant influence on TER. For the probability of going to university, after controlling for TER, the most significant school characteristics include the proportion on non-English speaking background students; the sector; and the socioeconomic make-up of the student body.

In this talk, Patrick will give an overview of the LSAY data, the research involved in producing this report and go into detail about the findings.

 Patrick Lim joined NCVER in 2007. He is a Senior Research Officer in the Research and Consultancy Branch of NCVER, but has also spent time in the Surveys, and Data Collection and Analysis branches, where he worked on apprentice and trainee completion rates. Prior to joining NCVER, he worked at Charles Sturt University as a lecturer in statistics, and as a statistical consultant for both the University of Adelaide and the Victorian Department of Primary Industries. Patrick has a background in mathematical statistics and experience in the application of statistical techniques to real-life problems. His areas of expertise include experimental design, linear mixed models and survey methodology.


Date: Wednesday October 16th

E. A. Cornish Memorial Lecture Announcement

South Australian Branch of the Statistical Society of Australia

Venue:    Napier G03, University of Adelaide, North Terrace. See the campus map, available at .

Time:   5.15pm        Pre-meeting drinks & nibbles (foyer).

6.05pm        E. A. Cornish Memorial Lecture.

7.30pm       Dinner will be held at Jasmin Resturant, 31Hindmarsh Square, Adelaide. RSVP to [email protected] by 2 October 2013.

Speaker:  Noel Cressie. Distinguished Professor, National Institute for Applied Statistics Research Australia (NIASRA), University of Wollongong. 

Topic: “Statistical Science: A Tale of Two Unknowns” 

Abstract: Science wants to answer the “Why” question and, in its pursuit, it often comes across the “Where” and “When” questions. Good data collection involves a protocol that specifies where (i.e., spatial locations) and when (i.e., temporal instants) the measurements were taken. Hence, it is possible to think about uncertainties in science as being partly explained by spatial variability and temporal variability. R. A. Fisher was very aware of spatial dependence in field trials; an externally imposed process of randomization was his way of dealing with it. E. A. Cornish advanced these ideas in Australian biometrics research. This talk weaves a tale of two unknowns, and I make a case that Fisher and Cornish would have found the Rev T. Bayes’ work highly relevant to their own if they had recognized “the other unknown.” 


Noel Cressie is Distinguished Professor in the National Institute for Applied Statistics Research Australia (NIASRA) and School of Mathematics and Applied Statistics, at the University of Wollongong. He has visiting positions at NASA’s Jet Propulsion Laboratory (JPL) and the Department of Statistics, University of Missouri. He received his Honours degree from the University of Western Australia and Master’s and PhD degrees from Princeton University. His research interests are in the statistical modelling and analysis of spatial and spatio-temporal data, and particularly in statistical applications to environmental data. He is the author of around 250 refereed articles and of three books, the most recent being “Statistics for Spatio-Temporal Data” by Noel Cressie and Christopher K. Wikle, published in 2011 by Wiley. Dr. Cressie is a Fellow of the American Statistical Association and of the Institute of Mathematical Statistics. Among other awards, in 2009 he received the COPSS Fisher Award and Lectureship.


Date: Wednesday September 18th, from 5:30pm

Speaker: Dr Olena Kravchuk, University of Adelaide.  

Topic: Random walks, Rank tests and R packages.


Would you agree with this, “if one wants to compare the means, but the data does not behave normally, one should use a nonparametric test, e.g. Mann-Whitney rank test, instead of the t-test” (a common recommendation in introductory statistics courses for biologists)?

The distribution-free property of rank tests only holds under their null hypotheses. The efficiency of rank procedures is governed by the data distribution. This feature of rank tests will be the subject of this talk. The talk corresponds to the way I teach statistics students Nonparametric Statistics.

There is an unassertive beauty and elegancy in rank tests that is completely missing in statistics teaching in non-specialised courses. In this talk, we will consider a random-walk visualisation of rank procedures and the behaviour of Hajek’s random walks under the alternatives of location and scale. I will discuss the practical appeal of such a visualisation as a tool for statistical data diagnostics and interpretations of rank tests. We will also take a closer look at the class of rank estimators associated with various rank tests. In particular, we will discuss situations in which efficient rank estimators are not (Pitman) consistent. I will show that such situations often tend to be ignored if one applies a straightforward computational approach to the rank data analysis.    

Olena Kravchuk is a senior lecturer in Applied Statistics in the School of Agriculture, Food and Wine, University of Adelaide. She has a strong research profile in applied statistics in Food Science, Sensory and Nutrition; she has published more than 40 journal papers and 3 book chapters in the last 7 years. She is a co-chair of the Australasian Applied Statistics Conference to be held in Port Lincoln in December, 2014. Olena’s recent research interest is in the theory and practical applications of rank tests for heavy-tail distributions. In particular, building on the efficiency of rank-estimators, she has proven the log-normality of the maximum likelihood estimator of scale for the Cauchy distribution (CommStat, 41(20): 3621-3632).


Date: Wednesday July 17th, from 5:30pm

Speaker: Lyron Winderbaum, University of Adelaide.  

Topic: Clustering of High-Dimensional Mass Spectrometry Data – and Applications to Ovarian Cancer.


Recent advances in Imaging Mass Spectrometry (IMS), extending the technology to the field of proteomics, have led to the acquisition of high-dimensional mass data from tissue sections of ovarian cancer. In such mass data, each sample corresponds to mass data from a point on the tissue section (with unique X-Y coordinates) and is represented by a curve or profile. These curves (functional data) are represented as high-dimensional vectors, and methods from multivariate analysis are therefore appropriate. Standard k-means clustering of these data produces spatially de-localized clusters which do not correspond with the histopathological annotation of the tissue section. I propose a transformation — including a smoothing step — of the original data and then apply a modified k-means algorithm to these transformed data which produces a number of clusters. These clusters are spatially localized, and match well with annotation made by the pathologist. Identifying the molecular discriminators of a cluster is of particular interest to the biochemists and clinicians and to this end I propose a dimension reduction step which results in a sparse representation for each sample. These sparse samples can then be visualized as a heatmap, which enhances the results of the cluster analysis as it leads to new and interpretable results. The sparse set of variables can be used to differentiate between datasets collected on tissue samples from different patients in a meaningful way. The hope is that such a differentiation could be used in the individualization of cancer treatments.

Lyron Winderbaum completed a B. Science (Chemistry) and a B. Math and Comp. Sci. (Pure Maths and Statistics) in 2010 before completing Honours in Statistics (applied topic in geology) in 2011. Lyron presented a paper at Young Stats Conference in Brisbane and gave a SSAI talk about analysis of a dataset on trace element concentrations in Pyrite (Fool’s Gold) from Moonlight prospect, Pajingo, a mine operated by Conquest Mining (formerly by North Queensland Metals (NQM)) in the Drummond basin. In 2012 Lyron began a PhD in Statistics and Biochemistry on the topic of this talk.

Date: Wednesday June 19th, from 5:30pm


Dr. Ian Saunders (CSIRO)


But what’s the R2?


Clients are used to seeing R2 values quoted for regression models and often request that we provide these for more complex models with multiple levels of variation and non-normal errors.

The meaning of R2 in these situations is not immediately clear and in fact a common response from a statistician is that there is no direct analogue to R2 so they should go away. This is not particularly helpful to the client.

The aim of this talk is to take another look at R2 and see if there is something more that we can offer. Possible definitions of R2 are examined in the context of a multi-stratum linear model and their practical implications discussed. We are not the first to consider this issue, but there seems to be a problem common to the writers on this subject in that they confuse the model with the data. Our aim is to make the situation a bit clearer by separating WHAT we are measuring with R2 from HOW we estimate it from data.

We conclude that there may still be a role for R2 in our toolkit as a useful and meaningful measure of the predictive power of a model.

This is joint work with Ros Miller and Tony Miller of CSIRO Mathematics, Informatics and Statistics.


Dr. Ian Saunders has had a long and somewhat chequered career in statistics. After a degree in Mathematics at Oxford and a Diploma in Mathematical Statistics and Cambridge, he moved to Australia to join CSIRO in 1974. He was initially based in Canberra, undertaking research and consulting in statistics and applied probability and obtaining a PhD in Statistics from the ANU. His work for CSIRO included modelling of Shakespeare’s vocabulary, the growth of small populations of Polynesians and estimation of rat populations without seeing any rats. In 1979 he moved to Melbourne, still with CSIRO, as a consultant in Building Research and Minerals, where he learnt SPlus and Unix, while analysing data on rotting timber and sampling iron ore from conveyor belts. He spent a year with the statistical consulting company Siromath and another at Colorado State University.

In 1988 Ian joined Richard Tweedie and others in setting up the statistics program and consulting centre at Bond University. This led him into Quality Management and in 1992 he was appointed Queensland Government Professor of Quality Management at the Queensland University of Technology, setting up a postgraduate teaching program in Quality. He was also a lead Evaluator for the Australian Quality Awards and Prize.

By 1995 he was itching to get back to analysis of real data and rejoined CSIRO in Adelaide, where after a period of commercial consulting in quality improvement and organisational performance measurement he joined the Bioinformatics group, being Program Leader for Quantitative Biosciences from 2008 to 2011. Most recently he has been working on studies in the Health area, finding a genetic link to inherited colon cancer susceptibility and also analysing data related to the CSIRO Total Wellbeing Diet.

He has been a member of the Statistical Society for the whole of his career, with periods on the ACT, Victorian and SA Branch committees, on the organising committees for the 6th and 15th Australian Statistical Conferences and on the SSAI Accreditation Committee, including 5 years as Chair.

Date: Wednesday May 15, 2013


Dr Murthy Mittinty (The University of Adelaide)


Imputing skewed continuous data: Advancement from the use of Normal imputation


With the availability of statistical software such as Proc MI in SAS or ICE in Stata, researchers are learning to deal with missing data and adjust for the bias due to missing respondents. Using either of the software researchers often imputes continuous data under an assumption of normality – yet many incomplete variables are skewed. We find that imputing skewed continuous variables under a normal model can lead to bias in the estimates. In this presentation we will be showing how Univariate skewed continuous variables under a normal model can lead to bias in the estimates. In this presentation we will be showing how Univariate skewed continuous variables can be imputed using more flexible distributions such as beta distribution and assess the extent to which these procedures work properly, and make comparisons with normal imputations. We will also discuss possible generalizations to the bivariate case using a special family of beta distributions.


Dr Murthy Mittinty has a PhD Statistics, from University of Canterbury, New Zealand. Murthy worked at National Institute for Water and Atmospheric Research, New Zealand dealing with the coastal flood and infectious disease modelling. He moved to Queensland, Australian and in 2008-2010 worked at Queensland University of Technology on projects relating to risk modelling in biosecurity. Since 2010 Murthy has been employed at the School of Population Health, University of Adelaide. His current work involves modelling of longitudinal data, missing data issues and also index development in nutritional epidemiology. Besides statistics Murthy has an interest in hiking, photography and cooking.

Date: Wednesday April 17, 2013


Dr Inna Kolyshkina (Institute of Analytics Professionals of Australia)


Selected Analytics industry case studies


At the launch of The Institute of Analytics Professionals of Australia (IAPA) Chapter in South Australia in September last year Dr Inna Kolyshkina gave a general talk on how analytics/data science added value to large organisations and provided some case studies and lessons learnt. There was interest from Statistical Society members in the methods used.

This talk will discuss in some detail technical aspects of the following recent industry case studies:

  • Bank mortgage default model
  • Telco call centre text record analysis to predict customer churn
  • Retailer – optimising ROI for television advertising

The talk will cover analytical process, data issues and resolution and the analytical techniques used.


Dr Inna Kolyshkina has 15 years’ experience delivering predictive analytics solutions for large organisations in Government, Banking, Retail, Telecommunications, Hospitality and Transport in Sydney.

Date: Wednesday March 20, 2013


2012 AGM Minutes

2012 Annual Report


Chris Davies

The University of Adelaide


A Comparison of Group-based Trajectory Modelling Methods


When an attribute is measured longitudinally in a population, sometimes the paths of measurements, or trajectories, that individuals follow are similar to one another. However, when certain characteristics are followed longitudinally, they show subpopulations with distinctly different trajectories.

Two main group-based trajectory modelling methods have been proposed to estimate these subpopulations, namely latent class growth analysis (LCGA) and growth mixture modelling (GMM). Both approaches are based on mixtures of groups following polynomial trajectories, with GMM providing the additional flexibility of hierarchical modelling within each group. Despite this, the LCGA model is widely used  and the consequences of its reduced flexibility have not been extensively explored. In both approaches an appropriate number of groups must be selected. This is usually achieved by assessing how a measure of model fit changes for different number of groups. A commonly used measure of model fit is the Bayesian Information Criterion (BIC).

Childhood behaviour trajectories may be amenable to modelling with group-based trajectory methods. We drew data concerning externalising childhood behaviours (e.g. aggression, bullying or delinquency) from the Generation 1 cohort, a prospective longitudinal study of South Australian children (n=557). To examine the suitability of these two approaches, a study was conducted using data simulated to match the characteristics of the Generation 1 sample. GMM was found to perform well for data simulated from both LCGA and GMM, whereas LCGA was satisfactory only for data simulated from LCGA. When the maximum BIC was used to select the number of groups, LCGA also overestimated the number of groups present in the data simulated from GMM.


Chris Davies is PhD Candidate in Statistics at The University of Adelaide and is also a Senior Statistician at the Data Management & Analysis Centre in the Discipline of Public Health of The University of Adelaide.

Date: Thursday November 22, 2012


Professor Richard Boys (University of Newcastle upon Tyne)


Inference for population dynamics in the Neolithic period


We consider parameter estimation for the spread of the Neolithic incipient farming across Europe using radiocarbon dates. We model the arrival time of farming at radiocarbon-dated, early Neolithic sites by a numerical solution to an advancing wavefront. We allow for (technical) uncertainty in the radiocardbon data, lack-of-fit of the deterministic model and use a Gaussian process to smooth spatial deviations from the model. Inference for the parameters in the wavefront model is complicated by the computational cost required to produce a single numerical solution. We therefore employ Gaussian process emulators for the arrival time of the advancing wavefront at each radiocarbon-dated site. We validate our model using predictive simulations. This work will shortly appear in a special issue of the Annals of Applied Statistics on the Mathematics of Planet Earth.


Richard Boys of Professor of Applied Statistics at the School of Mathematics & Statistics, Newcastle University, UK. He is currently on a research visit to the School of Mathematical Sciences at The University of Adelaide. Richard has been an active member of the Royal Statistical Society and is currently Chairman of their Graduate Training Programme Committee. He is also a member of the International Biometric Society, and the International Society for Bayesian Analysis. In 2008 Richard received a research grant from the Leverhulme Trust to study ‘Mathematical models for the developed Neolithic’. An upcoming edition of the Annals of Applied Statistics will publish a paper ‘Inference for population dynamics in the Neolithic period’. Richard’s main research interest is in the application of Bayesian Statistics to complex scientific problems through the use of computationally intensive Bayesian methods.

Date: Wednesday October 17, 2012


Professor Richard Jarrett (CSIRO)


Searching the parameter space for optimal conditions


This is a story about artifical corneal implants, which need to be “clear” (easy to see through) and also “permeable” (to allow nutrients to pass through). After two years of experiments, CSIRO scientists hadn’t managed to achieve both at once. I was asked to help design experiments to produce polymers that would achieve these two propertied simultaneously. A combination of fractional factorial experiments and response surface designs appeared to be just the ticket. But how to deal with the fact that there are two variables of interest? This talk will describe the approach we used – and our amazement when it worked! It says a lot about the “classical” training in working in high-dimensional vector spaces that was such a feature of the statistics courses at the University of Adelaide back in the days when the world was young.


Richard Jarrett is a Post-retirement Fellow at CSIRO Mathematics, Informatics & Statistics in Adelaide. He worked for two long stints for CSIRO in Melbourne, interspersed with periods as Professor of Statistics at the University of Adelaide and Director of the Statistical Consulting Centre at the University of Melbourne. He has been actice in the Statistical Society of Australia over many years, being President in 1989-90 and on the Accreditation Committee from its inception until 2003.

Date: Wednesday September 19, 2012

The Institute of Analytics Professionals of Australia (IAPA) is pleased to invite you to the launch event of our Chapter in South Australia. The launch event will be held on Wednesday 19 September at 5:30pm in Deloitte SA offices and will feature a presentation by Dr Inna Kolyshkina on ‘Analytics/data science adding value to large organisations. Cast studies and lessons learnt’.

When: September 19, 5:30pm for 6pm


Dr Inna Kolyshkina. Inna has 15 years’ experience delivering predictive analytics solutions for large organisations in Government, Banking, Retail, Telecommunications, Hospitality and Transport in Sydney.


Analytics/Data Science adding value to large organisations. Case studies and lessons learnt.

6:40pm onwards:

Informal networking session. Refreshments will be provided courtesy of our sponsors Deloitte and there will be ample opportunity to mix with local analytics practitioners.

Please find more information here:

IAPA events and membership do not involve any fees. IAPA prides itself in being vendor-neutral, so there are no sales pitches.

About IAPA:

Since 2004 IAPA has worked on uniting, informing and promoting analytics professionals, creating a networking hub and environment helping to share best industry practices. There are 2000+ IAPA members Australia-wide. Please see for a snapshot of our membership by industry. IAPA Chapters in NSW, Vic, Qld, SA, WA and ACT hold quarterly events, support local analytics conferences and education initiatives. The SA Chapter of IAPA aims to do the same, tailoring the activity to the South Australian market specifics.

Date: Thursday August 16, 2012


Professor Jonathan Karnon (The University of Adelaide)


Taking stock in health care: applying economics, statistics and business processes to reduce waste and improve patient outcomes


Over the last five years, the Adelaide Health Economics Research Group (AHERG) has been developing and applying methods for the economic evaluation of applied clinical services. The focus is on the use of linked administrative data to inform clinical and policy actions to reduce important variation in clinical practice, where ‘important’ refers to variations in clinical practice that result in significant differences in the cost-effectiveness of service provision.

A two-stage evaluative framework has been developed, involving:

  • risk adjusted cost-effectiveness (RAC-E) analyses of alternative applied services to identify clinical areas with important variation (the risk adjusted refers to adjustments for varying levels of baseline risk for increased costs or poor outcomes)
  • analyses of clinical practice processes at the alternative service providers (eg, hospitals) to identify key areas of difference that may be associated with the RAC-E results from stage 1.

Presented to stakeholders, we hypothesise these data will incentivise and support actions to reduce important variation in clinical practice processes, as well as enabling ongoing monitoring of the impact of performance improvement initiatives.

To date, we have undertaken RAC-E analyses of hospital-based services for patients presenting with chest pain, stroke, hip fracture, and amputations across the four main public hospitals in South Australia. We have also applied process mining techniques to describe clinical practice processes for patients attending the emergency departments (EDs) of the four main public hospitals in South Australia with chest pain. Process mining is a relatively new area of research, which combines analyses of the content and sequential order of components of the clinical process (workflow analyses), with analyses of the time between key events (performance analyses).

In this talk, I will present the methods and results of the RAC-E and process mining analyses of the alternative clinical practices processes for patients presenting with chest pain at the four main public hospitals in South Australia. In particular, I will focus on the methods used for risk adjustment, for which we have used ‘genetic matching’ (a method of multivariate matching, that uses an evolutionary search algorithm to determine the weight each covariate is given), propensity-weighted regression-based adjustment, and non-weighted regression-based adjustment.


Jonathan Karnon is Professor in Health Economics at the Adelaide Health Economics Research Group (AHERG), University of Adelaide. John is convenor of the postgraduate Health Economics program and within that program he delivers a course in Health Economic Evaluation and Decision Making. John has a wide range of experience in both applied and methodological research. His principle focus is around the use of decision modelling techniques as a framework for the economic evaluation of health care technologies, and more recently the use of linked, routinely collected data to inform the economic evaluation of existing services.

Recent papers:

Karnon J, Caffrey O, Pham C, Grieve R, Ben-Tovim D, Hakendorf P, Crotty M. Applying risk adjusted cost-effectiveness (RAC-E) analysis to hospitals: estimating the costs and consequences of variation in clinical practice. Health Economics 2012; DOI: 10.1002/hec.2828

Karnon J, Ben-Tovim DI, Pham CT, et al. The efficient price: an opportunity for funding refrom. Australian Health Review 2011; 35(4): 501-506

Date: Wednesday June 20, 2012


Shahid Ullah (Flinders University)


Accurate forecast using functional data analysis


Time series forecasting is an important research problem for many real life domains such as forecasting fertility and mortality rates. There have been many methodological developments in demographic forecasting during the last two decades. It is important that good statistical approaches are used to generate accurate and reliable information about future data to inform public health investment decisions. It is critical, therefore, that such forecasting are derived using the best avilable statistical approach to minimize possible errors in the future. A new method is proposed for forecasting age-specific mortality and feritility rates observed over time. This approach allows for

  1. smooth functions of age (useful graphical tools),
  2. is robust for outlying years (due to wars and epidemics),
  3. cohort effects,
  4. prediction intervals, and
  5. provides a modelling framework that is easily adapted to allow for constraints and other information.

Ideas from functional data analysis, nonparametric smoothing and robust statistics and combined to form a methodology that is widely applicable to any functional time series data observed discretely and possibly with error. The methodology is illustrated by using Australian demographic data.


Dr Shahid Ullah is a Senior Lecturer in Biostatistics for Flinders Centre for Epidemiology and Biostatistics. His responsibilities include statistical consulting to staff and postgraduate students within the Faculty of Health Sciences in addition to teaching and independent research. Whilst he has an excellent broad-based research background and expertise in the field of applied statistics in the last 15 years, his broad research interests are now focused on functional, epidemiological and statistical model building in the area of health sciences.

Shahid received his PhD degree in Statistics from Monash University and his PhD work provided a new and innovative set of tools known as functional data analysis that enable any time series modelling and forecasting with uncertainty. Shahid contributed to a strong group of researchers in sports injury prevention within the University of Ballarat by providing expert consulting advice as well as leading his own statistical projects. Shahid has collaborated widely on numerous cross-institutional projects and his research activities have been disseminated through 72 authored publications in peer-reviewed journals, reports, conference presentations and other papers. He has been awarded over $1.0 million in collaborative research grants.

Date: Wednesday May 16, 2012


Graeme Tucker (SA Health)


Planned home births in South Australia, 1991-2006: Lessons now and for the future


The objective of the study was to examine diferences in outcome between planned home births, occuring at home or in hospital, and planned hospital births. A retrospective population-based study using South Australian perinatal data on all births and perinatal deaths 1991-2006 was conducted, from which both planned home and hospital births can be used to assess current practice and outcomes. The study provides a baseline for evaluating the impact of the ‘Planned Birth at Home’ policy and changes in the types of maternity care available to pregnant women.

The main perinatal outcomes studied were perinatal death, intrapartum death, death attributed to intrapartum asphyxia, Apgar score <7 at five minutes, and use of specialised neonatal care or paediatric intensive care. Maternal outcomes assessed were postpartum haemorrhage and perineal injury after vaginal birth.

Analysis included logistic regression adjusted for predictor variables and standardised mortality ratios.


Graeme Tucker is a research statistician with 37 years experience in the field. Graeme worked fro the Australian Bureau of Statistics for 17 and a half years, 7 of the last 8 years in Statistical Services providing statistical consultancy and advice. He then worked for the State Department for Family and Community Services for 5 years until the restructure of the Department, providing statistical and evaluation services and consultancy to the Policy Division. This Department formed a part of the new Department of Human Services, until the Department was split into Health, and Family and Community Services. During this process Graeme moved to the Epidemiology Branch of the Health Department, where he has provided similar statistical and consultancy services for the past 11 years, both to the Department of Health and other government agencies.

Graeme has also been active in providing statistical services in the private sector for the last 25 years.

Date: Wednesday April 18, 2012


A/Prof Richard J Woodman (Flinders University)


Assessment of model performance in disease risk prediction; out with the old and in with the new?


Quantification of the added usefulness of new measures in risk prediction has traditionally relied upon significance tests from regression models and increases in the C-statistic. However, significance model predictors often cause only minor increases in the C-statistic suggesting limited utility of the new measures in improving risk prediction. More recently, other discriminators have gained in popularity amongst researchers. The Integrated Discrimination Improvement index (IDI) measures the difference beteween the change in the mean predicted risk of an event occurring for those who had the event, and the change for those who didn’t ahve the event. The Net Reclassification Improvement index (NRI) quantifies the percentage of subjects correctly re-classified in terms of risk. In this talk I provide an example that compares results and conclusions from each of these measures.

A logistic regression model was used to predict risk of long from short (≤ 72 hrs) hospital stay amongst 1457 general medicine patients. Significant predictors were age, blood pressure (BP), heart rate (HR), respiratory rate (RR), mobility, white blood cell count (WBC), cardiac failure (CF) and the need for supplemental oxygen (SuO2). Using the predicted probabilities for long-stay, we assessed improvements in the C-statistic (∆ C), the IDI (%) and the NRI (%) after the addition of each variable beyond age. The NRI was assessed using predicted probability cut-points for long-stay of 50% and 57% (ie the overall prevalence of long-stay patients), and the category-free NRI, that assesses the proportion of patients with improved prediction probabilities according to their eventual outcome.


Richard Woodman is Director of the Flinders Centre for Epidemiology and Biostatistics (FCEB), in the Discipline of General Practice, Flinders University.

E. A. Cornish Memorial Lecture

Date: Wednesday November 16, 2011


Peter Diggle (Lancaster University)


A Tale of Two Parasites: Model-based Geostatistics and River Blindness in Equatorial Africa


I will begin this lecture with some general remakes on the role of statistics in scientific research, as my own approach to research has been greatly influenced by my time with CSIRO and the culture that Cornish engendered in his time as first chief of the Division of Mathematical Statistics.

For the remainder of the lecture, which I hope will deomonstrate this approach in action, I will give an introduction to model-based geostatistics and describe its contribution to the African Programme for Onchocerciasis Control. This multi-national programme aims to control onchocerciasis (river blindness) throughout the affected region of Africa by widespread distribution of a filaricide medication, ivermectin. In areas co-endemic for river blindness and Loa loa (eye-worm), the programme has been interrupted by the occurrence of severe, occasionally fatal, reactions to ivermectin. This has necessitated the mapping of eye-worm prevalence to identify potentially high-risk areas, using (geo) statistical modelling to combine the information provided by spatially sparse prevalence surveys and spatially dense environmental covariates.

Crainiceanu, C. Diggle, P.J. and Rowlingson, B.S. (2008) Bivariate modeling and prediction of spatial variation in Loa loa prevalence in tropical Africa (with Discussion). Journal of the American Statistical Association, 103, 21-47.

Diggle, P.J., Thomson M.C., Christensen, O.F., Rowlingson, B., Obsomer, V., Gardon, J., Wanji, S., Takougang, I., Enyong, P., Kamgno, J., Remme, H., Boussinesq, M. and Molyneux, D.H. (2007). Spatial modelling and prediction of Loa loa risk: decision making under uncertainty. Annals of Tropical Medicine and Parasitology, 101, 499-509.

Diggle, P.J. and Chetwynd, A.G. (2011). Statistics and Scientific Method: an Introduction for Students and Researchers. Oxford: Oxford University Press.

Zoure, H., Wanji, S., Noma, M., Amazigo, U., Diggle, P.J., Tekle, A. and Remme, J.H. (2011). The geographic distribution of Loa loa in Africa: results of large-scale implementation of the Rapid Assessment Procedure for Loiasis (RAPLOA). Public Library of Science: Neglected Tropical Diseases (to appear).


Peter Diggle is Distinguished University Professor of Statistics in the School of Health and Medicine, Lancaster University, Adjunct Professor in the Deparatment of Biostatistics, Johns Hopkins University School of Public Health and Adjunct Senior Researcher in the International Research Institute for Climate and Society, Columbia University.

Between 1974 and 1983 Peter was a Lecturer, then Reader, in Statistics at the University of Newcastle upon Tyne. Between 1984 and 1988 he was Senior, then Principal, then Chief Research Scientist and Chief of the Division of Mathematics and Statistics and CSIRO, Australia.

Peter’s research interests are in the development of statistical methods for spatial and longitudinal data analysis, motivated by applicatios in the biomedical, health and environmental sciences. he has published 10 books and around 180 articles on these topics in the open literature. He was awarded the Royal Statistical Society’s Guy Medal in Silver in 1997, is a former editor of the Society’s Journal, Series B and is a Fellow of the American Statistical Association.

Peter was founding co-editor, with his close friend and Johns Hopkins colleague Scott Zeger, of the journal “Biostatistics” between 1999 and 2000. He is a trustee for Biometrika, and has served the UK Medical Research Council as a member of their Population and Systems Medicine Research Board.

Away from work, Peter plays mixed-doubles badminto with his family (partner Amanda, children Jono and Hannah). He is a keen cook, and should play more badminto than he does to counteract the effects of this. He also enjoys music – playing guitar and tenor recorder, and listening to jazz.


Dinner at Jah’z Lounge, 7 Cinema Place, Adelaide will consist of a $40 per head fixed price menu, consisting of an entree of tapas style platters and then a main course selected from a steak dish, a chicken dish and a vegetarian dish. RSVP to Paul Sutcliffe, [email protected], by 2 November 2011, and please note that places are limited.


Also during his visit to Adelaide Peter will be presenting a two-day workshop on ‘Model-based geostatistics: with applications in the environmental and health sciences’, at Flinders University. Please see here for more details and registration.

Joint SSAI SA branch / Australasian Epidemiology Association (AEA) meeting

Date: Wednesday October 19, 2011


John Lynch (University of Adelaide)


The SANT Data Link Early Childhood Demonstration Project: the experience so far


John Lynch is Professor of Public Health at the University of Adelaide since early 2011. He is also Visiting Professor of Epidemiology at University of Bristol (UK). He was previously in the Department of Epidemiology at the University of Michigan (USA) and was a Canada Research Chair in the Department of Epidemiology and Biostatistics at McGill University in Montreal (Canada). In mid 2008 he returned to Australia and took up an appointment at University of South Australia. He is an internationally recognised scholar in epidemiology and public health with more than 200 publications. In 2007 his work in public health was recognised with an honourary Doctorate in Medical Science from the University of Copanhagen. In 2009 he was awarded a prestigious NHMRC Australia Fellowship.

His research interests include early childhood development, life course processes regulating health behaviours and human capability formation, population health information systems, evidence-based public health and improving the public health research-policy nexus.

Current research projects:

  • Using linked data to monitor population health and interventions
  • Predictive validity of routinely collected perinatal data for child development
  • Understanding population trajectories of healthy child development
  • Efficacy evaluation of a nurse-lead home visiting program
  • Internet-based parental support for infant and child development
  • Trajectories of childhood growth and adolescent risk factors in the Aboriginal Birth Cohort
  • Cognitive and non-cognitive characteristics, child development, health behaviours and social trajectories of human capability formation
  • Evidence for interventions to reduce inequalities in child healthy development
  • Role of weaning diet on adolescent cognitive function and physiological risk factors

SSAI Student Talks

Date: Thursday September 22, 2011


Lyron Winderbaum (University of Adelaide)


Fool’s Gold – It’ll fool you


I investigate a small data set on trace element concentrations in Pyrite (Fool’s Gold) from Moonlight prospect, Pajingo, a mine operated by Conquest Mining (formerly by North Queensland Metals (NQM)) in the Drummond basin, North Queensland. The financial motivation is primarily prediction of gold concentration. Several multivariate methods of explorative analyses are considered, including cluster analysis of the trace elements by the psuedo-metric 1-abs(cor(X1,X2)). The cluster analysis dendrogram is shown to nicely encapsulate the information contained in the pairwise correlations. Information provided by other methods is compared to it. I then demonstrate how, due to the nature of the data, such methods can potentially be extremely misleading, as the correlation coefficients are so strongly affected by outlying values. The effect of outliers is greatly reduced by dealing with the logarithm of the concentrations, but the original clear trends are lost. The well established relationship between Gold and Arsenic in pyrite is a known method for prediction of gold. More complex models are considered and compared for improvement in prediction of Gold concentration.


Aiden Fisher (University of Adelaide)


The Phase-type distribution – the forgotten distribution?


The phase-type distribution is a versatile, but often overlooked, probability distribution. The continuous univariate form can be used to approximate any continuous distribution with non-negative support, with weak convergence as the number of phases increases. Despite its versatility it is notably absent in more of the statistical literature, and rarely appears outside the reliability context. After a brief discussion on its characterisation and its natural relationship to Markov theory, two practical applications for its use in survival analysis and time-series analysis are given.

Date: Wednesday July 20, 2011


Lisa Yelland is a Senior Statistician at the Australian Research Centre for Health of Women and Babies (ARCH) and at the Data Management and Analysis Centre (DMAC), both at the University of Adelaide.


Analysis of Perinatal Trials Including Multiple Births: When Should Clustering be Taken Into Account?


Perinatal trials are used to determine the effect of particular treatments or interventions on the health and wellbeing of mothers and their babies. In the simplest case, babies in a perinatal trial will be independent and the data can be analysed using standard statistical methods, such as linear or logistic regression. Clustering occurs when babies from a multiple birth (e.g. twins) are included in the trial. The outcomes of babies from the same birth are likely to be related due to shared genetic and environmental factors. This violates the assumption of independence that is required for standard statistical methods to be valid. Methods which account for clustering are available; however, these are generally only used in settings where all subjects are part of a cluster (e.g. twin studies). Conflicting recommendations have been made regarding if and when clustering should be taken into account in the analysis of perinatal trials including infants from both single and multiple births, particularly when the multiple birth rate is low.

In this talk, I explore some of the factors that influence whether clustering due to multiple births should be taken into account in the analysis of perinatal trials. The DINO Trial, in which preterm infants received either high doses or standard doses of fish oil, is used as an illustrative example throughout. The results of a large simulation study show that standard statistical methods which treat all babies as independent perform poorly in some settings. Other methods which account for the clustering can give more appropriate results. I make recommendations for analysing data from perinatal trials including infants from multiple births.


Lisa Yelland is a Senior Statistician at the Australian Research Centre for Health of Women and Babies (ARCH) and at the Data Management and Analysis Centre (DMAC), both at the University of Adelaide. She recently submitted a PhD on statistical methods for analysing randomised controlled trials with a focus on applications to the perinatal trial setting.

Date: Wednesday June 15, 2011


Professor Corey J. A. Bradshaw, Director of Ecological Modelling, The Environment Institute and School of Earth & Environmental Sciences, The University of Adelaide


Statistical and simulation models estimating extinction time from fossil evidence: climate-driven megafaunal turnover in the Late Pleistocene


The rate at which a once-abundant population declines in density prior to extinction can strongly influence the precision of statistical estimates of extinction time. Here we report the development of a new, robust method of inference which accounts for these potential biases and uncertainties, and test it against known simulated data and dated Pleistocene fossil remains (mammoths, horses and Neanderthals). Our method is a Gaussian-resampled, inverse-weighted McInerny (GRIWM) approach which weights observations inversely according to their temporal distance from the last observation of a species’ confirmed ocurrence, and for dates with associated radiometric errors, is able to sample individual dates from an underlying fossilization probability distribution. We show that this leads to less biased estimates of the ‘true’ extinction date. In general, our method provides a flexible tool for hypothesis testing, including inferring the probability that the extinctions of pairs or groups of species overlap, and for more robustly evaluating the relative likelihood of different extinction drivers such as climate perturbation and human exploitation. We show the application of this approach using genetic data from 25 detailed time series of regional extinctions and invasions across the Northern Hemisphere over the past 60,000 years.


Professor Corey Bradshaw currently holds the Professorial Chair of Ecological Modelling at the University of Adelaide’s Environment Institute, and he has a co-appointment with the South Australian Research and Development Institute. Previously, he held an Associate Professor position at Charles Darwin University in northern Australia, and was a postdoctoral fellow at the University of Tasmania. Corey has completed three tertiary degrees in mathematical ecology (BSc, MSc, PhD) from universities in Canada and New Zealand.

Corey believes that to appreciate the impact of deforestation, pollution, habitat loss, extinctions, over-grazing, over-fishing or warming climates on human wealth, health and well-being, conservation ecologists must analyse information from different ecosystems worldwide to estimate the impact of human activity on all life forms. He specialises in applying mathematical models to large multi-species datasets to determine global-scale patterms of threat to biodiversity. In a world where human activity has precipitated the current Anthropocene extinction event and the subsequent loss of hundreds of thousands of species, his aim is to provide irrefutable evidence to influence government policy and private behaviour for the preservation of functioning ecosystems.

Corey has published over 150 peer-reviewed scientific articles, 8 book chapters and a book. He is a fellow of the Royal Society of South Australia and was awarded the 2010 Australian Ecology Research Award, the 2010 Scopus Young Researcher of the Year, the 2009 HG Andrewartha Medal, and a 2008 Young Tall Poppy Science Award. He is also editor of two international scientific journals and is regularly featured in Australian and international media for his research.

Date: Wednesday May 18, 2011


Dr Julian Taylor, CSIRO


Variable Selection in Linear Mixed Models with application to QTL analysis


One common focus in modern plant breeding experiments is the analysis of Quantitative Trait Loci (QTL). Unfortunately these experiments are often complex requiring non-genetic sources of variation to be captured such as structured spatial correlations between observations and/or variation arising from other experimental design components. This talk will discuss the theory and application of whole genome QTL analyses using a variable selection method integrated into current mixed model theory. The integration is achieved through an extension and reparameterisation of a well known class of penalties. To ensure efficiency the Mixed Model Variable Selection (MMVS) method uses the flexible software package ASReml-R (Butler, et al. 2009) as its core linear mixed model fitting routine. A general simulation study reveals the extended class of penalities achieves varying degress of estimator shrinkage depending on one of its parameters. The simulations also reveal a link between the number of false positives and the number of true coefficients using the same penalty. The MMVS method is then applied to a wheat quality data set from the Food Futures Flagship, CSIRO where the focus is the analysis and interpretation of QTL.


Julian Taylor is a postdoctoral fellow at the Mathematics, Informatics and Statistics division of CSIRO where he is affiliated with the Food Futures National Research Flagship. He currently collaborates on projects with Plant Industry including the analysis of QTL for wheat quality with advanced intercrosses.

Date: Wednesday April 20, 2011


Caroline Deans, Assistant Director, South Australian Census Management Unit, Australian Bureau of Statistics


2011 Census: 111 days to go


The Australian Census of Population and Housing will be conducted on 9 August 2011. The Census is often described as Australia’s largest peace time operation, costing $440 million over its five year cycle and employing 43,000 people. It is run like a military operation with clean lines of command, well defined procedures to ensure consistency and a massive logistical exercise to transport 1,400 tons of Census material around the country.

The presentation will discuss:

  • Why Australia takes a Census? Who uses the data and what is it used for?
  • How is the Census conducted? Including the enumeration methodology, post enumeration survey and processing of data.
  • Strategies to reduce undercoverage in hard to reach population groups


Caroline Deans is an Assistant Director at the ABS responsible for the South Australian Census Management Unit. Caroline has been involved in the Census program for 18 months, successfully managing the 2010 Dress Rehearsal and now overseeing a team of 25 to conduct the 2011 Census in South Australia. Caroline joined the ABS five years ago after a career in the private sector, including stints at the RAA and ETSA. She has a Masters of Business Administration and a Bachelor of Management (International Marketing). Caroline has lived in Adelaide for most of her life and is getting to know country South Australia better as she travels the state recruiting staff and engaging with local stakeholders.

Date: Wednesday March 23, 2011


2010 AGM Minutes

2010 Annual Report


Dr Chris Brien

University of South Australia


Principles in the design of multiphase experiments with a later laboratory phase: orthogonal designs


It is common for the material produced from field and other experiments to be processed in a laboratory. Reasons for this include the need to measure chemical and physical attributes using equipment such as spectrometers, gas chromatographs, pH meters or wear and strength testers, or to produce processed products such as wine, bread and malt that are subsequently assessed, often by an expert panel. These experiments are multiphase. They occur widely in agriculture, food processing and pharmaceutical industries and biological and environmental sciences, although their multiphase nature is often not recognised.

A systematic approach to designing the laboratory phase of such multiphase experiments, taking into account previous phases, will be described. We extend the fundamental principles fo experimental design – randomization, replication and blocking – to provide general principles for designing multiphase experiments that employ orthogonal designs. In particular, the need to randomize the material produced from the first phase in the laboratory phase is emphasized. Factor-allocation diagrams are employed to depict the randomizations in a design and skeleton analysis-of-variance (ANOVA) tables to evaluate their properties. The techniques are applied to several scenarios for an athlete training experiment.


Dr Chris Brien is currently an Adjunct Senior Lecturer in the Phenomics and Bioinformatics Research Centre at the University of South Australia. Chris, while principally a Biometrician, has taught and consulted in many application areas of statistics. His research interest is in the design and analysis of experiments. In particular he has been a pioneer in the design and analysis of multiphase and other multitiered experiments. Another area of interest has been the formulation of mixed models for experiments, including for longitudinal experiments. Has has been a GenStat user since 1971 and these days also uses ASReml-R.

Date: Thursday February 17, 2011


Liliana Orellana

Chair, Institute of Calculus, University of Buenos Aires


Estimating the optimal dynamic treatment regime from longitudinal observational data


Dynamic treatment regimes are individually tailored treatments based on patient covariate history. Optimal dynamic regimes (ODR) are rules that will lead to the highest expected value of some utility function at the end of a time period. Many pressing public health questions are concerned with finding the ODR out of a small set of rules in which the decision maker can only use a subset of the observed covariates. For example, one pressing question in AIDS research is to define the optimal threshold CD4 cell count at which to start prescribing HAART to HIV infected subjects, a rule which only depends on the covariate history through the minimum CD4 count.

In the last decade there have been important advances in the development of methodology adequate to estimate optimal treatment strategies from observational longitudinal cohorts. In this talk I will present one approach to estimate the ODR when the set of enforceable regimes comprises simple rules based on a subset of past information which can be indexed by a Euclidean vector x. In will describe how to conduct inference, using an inverse probability weighting approach, under models that allow the possibility of borrowing information across regimes and across baseline covariates. Finally I will point out some of the practical issues that remain unresolved for the specific problem of estimating the optimal CD4 cut-off point at which to start HAART therapy.

This is joint work with Andrea Rotnitzky and Jamie Robins.


Liliana Orellana is chair of the Institute of Calculus at the University of Buenos Aires. The Institute is a research centre focusing on Statistics and Applied Mathematics. Liliana is the leader of the Graduate Diploma in Statistics for Health Sciences (University of Buenos Aires), a two year graduate program intended for healthcare professionals with a strong emphasis on applied statistics. She is also an Assistant Professor at the Faculty of Sciences where she lectures statistics courses at both undergraduate and graduate levels. Her current research focuses on methods for drawing causal inferences from longitudinal studies. Liliana has contributed with original methodology for estimating the optimal treatment regime from longitudinal data collected in an observational study. The approach was aimed at elucidating the optimal time to start antiretroviral therapy in HIV infected persons.

 Date: Tuesday November 9, 2010


Prof R. A. Bailey

Professor of Statistics, Mathematical Sciences Institute, Queen Mary University of London


Conflicts between optimality criteria for block designs with low replication


The quality of incomplete-block designs is commonly assessed by the A-, D-, and E-optimality criteria. If there exists a balanced incomplete-block design for the given parameters, then it is optimal on all these criteria. It is therefore natural to use the proxy criteria of (almost) equal replication and (almost) equal concurrences when choosing a block design.

However, work over the last decade for block size 2 has shown that when the number of blocks is near the lower limit for estimability of all treatment contrasts then the D-criterion favours very different designs from the A- and E-criteria. In fact, the A- and E-optimal designs are far from equi-replicate and are amongst the worst on the D-criterion.

I shall report on current work which extends these results to all block sizes. Thus the problem is not blocks of size 2; it is low replication.


Rosemary A. Bailey is a British statistician who is renowned for her work in the design of experiments and the analysis of variance and in related areas of combinatorial design, especially in association schemes. She has written books on the design of experiments, on association schemes, and on linear models in statistics. She is a professor of statistics at the Mathematical Sciences Institute, Queen Mary University of London, England.

Date: Wednesday October 20, 2010


Charles Pearce

Thomas Elder Chair of Mathematics, The University of Adelaide


Statistics and New Zealand


Prehistory commonly proceeds in terms of scenarios proposed by historians. These scenarios often change dramatically in the light of new discoveries. Genetics in particular has had an important role in the last decade in bringing about such change. This talk gives an example of the power of statistics to overturn prevailing historical paradigms.

Recently there has been a growing acceptance of a late (12th century AD) first settlement date for New Zealand by Polynesians. This is sometimes advanced to AD 1280. For a modest initial number of settlers, a very rapid and sustained population growth is then required to produce the population extant around AD 1800. We indicate how statistics coupled with palaeodemography establishes that this paradigm, although currently dominant, is untenable, and that the evidence requires a much longer time frame for New Zealand prehistory.

Reference: “Oceanic Migration”, Charles and Frances Pearce. Published by Springer 2010.


Charles Pearce was born in New Zealand. After completing degrees in mathematics and physics at Wellington, he obtained a Ph.D. in the Department of Mathematical Statistics at the ANU. His research interests centre on probabilistic and statistical modelling and analysis and optimization, in which he has over 300 publications. He was awarded the ANZIAM Medal in 2001 and the Potts Medal in 2007 for outstanding contributions to applied mathematics and operations research. He holds the Thomas Elder Chair of Mathematics at The University of Adelaide.

Date: Wednesday September 15, 2010


Tony Meissner

Principal Scientist – Monitoring, Resources Monitoring Team, Division of Science, Monitoring and Information, Department for Water at Berri, South Australia


Statistical Models of Impact of Woolpunda Salt Interception Scheme on the River Murray


The River Murray as it flows through South Australia acts as a drain for the highly saline regional groundwater that contributes to salinity in the river. The largest input of salt, approximately 250 t/d, occurs between Lock 3 and Holder downstream. Forty nine bores, above the river valley, were drilled to the underlying groundwater from 1989 to 1992. Pumping of saline ground-water commenced in 1990 to lower the groundwater gradient to the river thus preventing salty water entering the river. The Woolpunda Salt Interception Scheme became fully operational in June 1996. A hydrogeological review of the scheme was carried out in August 2000 (Telfer and Way, 2000) that concluded the impact of the scheme for flows less than 10,000 Ml/d was -46.3 EC units at Morgan, SA.

Daily readings of salinity (EC) and flow are taken at Overland Corner 14 km downstream from Lock 3 and salinity at Holder, 26 km downstream from Overland Corner. Weekly average readings were calculated from the daily values from April 20 1992 until Dec 31 1999. A factor (phase) was derived from the drawdown period from April 1992 to June 1996 and the operating period from July 1996 to December 1999. Three statistical linear mixed effects models were examined: (1) the EC value at Holder was regressed against the EC and flow at Overland Corner and phase (2) The difference in EC value between Holder and Overland Corner regressed against flow and phase, and (3) the two sites were converted to factors (sites) and EC regressed against flow, phase and sites. At flows of 5,000 Ml/d the impact of the SIS was estimated, for the three models, to be 30:2 +/- 10:1, 34:6 +/- 3:4, and 32:4 +/- 2:3 EC units respectively.


Tony Meissner currently works as Principal Scientist – Monitoring in the Resources Monitoring Team, Division of Science, Monitoring and Information, Department for Water at Berri, South Australia. He has had over 40 years experiences in agricultural and water resources research and management. In 2005 he spent a year with the Murray-Darling Basin Commission, Canberra contributing to salinity policy. On his return, he managed the Berri Hydrometric Unit who undertake flow and salinity monitoring of the River Murray. In the past 18 months Tony has contributed to the analysis of hydrological data particularly along the River Murray in South Australia.

Date: Wednesday August 18, 2010

SPECIAL EVENT – Joint meeting with the Australasian Epidemiological Association


Dr Emily Steele (Research Fellow) & Dr Lynne Giles (Senior Research Fellow)

Life Course and Intergenerational Health Research Group, Robinson Institute, The University of Adelaide


The influence of precarious employment on age at first childbirth: some epidemiological and statistical considerations


The mean age at which women in Australia have their first child increased in recent decards to reach 28.2 years by the year 2006. Since maternal age is a strong risk factor for infertility, pregnancy complications and neonatal problems, it is imperative to investigate barriers to women having children when they (and their partners) would like to. We aimed to assess women’s (and their partners’) experiences of precarious employment as a factor in older age of first childbirth. We concomitantly considered women’s educational attainment, and the influence of other contemporary sources of financial insecurity, such as having a higher education debt.

We conducted a cross-sectional retrospective study based on a birth cohort of young women (n ~ 1000, born 1973-75). A detailed event history instrument was developed to obtain data regarding a range of life domains over a 20 year life course period,  including pregnancy, partnering, education, and employment histories. Much of the data was collected at the the month-level of detail. Time-varying and time-constant survival analysis techniques were applied within a life course framework to examine the effects of precarious employment on age at first childbirth (takingi nto account educational attainment and other influential factors), with a sub-set of the study cohort (n=230). This presentation will address key challenges we experienced in designing the analytical framework.


Emily Steele is a Research Fellow in the Life Course and Intergenerational Health Research Group in the Robinson Institute at The University of Adelaide. She recently finished her PhD, which focused on the influence of life course structural determinants of older first childbirth and included the design of complex epidemiological survey instruments, and data collection (and analysis) from a population-based cohort (n ~ 1000). Emily is a physiotherapist and also completed a Master of Public Health prior to her PhD.

Lynne Giles is a Senior Research Fellow in the Life Course and Intergenerational Health Research Group in the Robinson Institute at The University of Adelaide. Her interests include social epidemiology and the analysis of longitudinal data. After gaining her undergraduate qualifications in Mathematical Statistics, Lynne completed a Master of Public Health and PhD in Applied Statistics.

 Date: Wednesday July 21, 2010

OPEN FORUM – The South Australian Branch of the Statistical Society in conjunction with the Mathematicians in Schools program invite you to an open forum to hear about and discuss ways to encourage more students to study mathematics and ways of improving the teaching of mathematics and statistics in Australian schools and universities.


Dr Rebecca Anderson

SA Project Officer, Mathematicians in Schools, CSIRO Education


Mathematics in Schools Program


Mathematicians in Schools aims to create and support long-term professional partnerships between mathematicians and teachers. Its purpose is to promote a deeper understanding of the importance of mathematics in our society for students and teachers, and through them, the wider community. Mathematicians in Schools in an Australian Government initiative that is managed by CSIRO Education.

Open forum:

Paul Sutcliffe (SA Branch President) will chair an open forum aimed at identifiying ways to improve the number of students studying higher level mathematics. How can we encourage those students studying mathematics at University to consider statistics as an option?

Date: Wednesday June 23, 2010

SPECIAL EVENT – Joint meeting with the Australasian Epidemiological Association


Dr Nicole Pratt

Senior Research Fellow, Sansom Institute, University of South Australia


Using an Australian administrative data set for post-marketing surveillance of antipsychotics in elderly veterans: The challenge of unmeasured confounding


Computerised administrative claims databases provide a convenient and valuable source of information to study the effects of medicine use, however, observational studies utilizing these data are often criticized due to the potential lack of control for unmeasured confounding. The extent to which traditional pharmacoepidemiological studies utilising administrative claims databases can deal with confounding is limited as these data sources often lack information on many potentially important confounders, such as clinical information, life style factors and disease severity.

In this talk, I will introduce some new approaches to help overcome possible bias in observational studies due to unmeasured confounding, including: instrumental variable analysis, the self-controlled case-series design, prescription sequence symmetry analysis and propensity scores. I will then illustrate how these approaches apply to the assessment of the adverse effects of antipsychotic medication prescribing in the elderly.


Nicole is a senior research fellow in the Quality Use of Medicines and Pharmacy Research Centre, Sansom Institute, University of South Australia. Nicole has recently completed her PhD entitled ‘Medication prescribing in the elderly and the effect on health related outcomes: An investigation of bias in observational studies using computerised claims databases’. The aim of this thesis was to determine how information contained in computerised administrative claims databases can be used effectively to study the outcomes of medicine use. This work focused on the techniques and methods that are currently available to pharmacoepidemiologists to overcome unmeasured confounding. Nicole has worked previously as a biostatistician with the Data Management and Analysis Centre, in the Discipline of Public Health, University of Adelaide.

Date: Wednesday April 28, 2010


Dr Jessica Kasza

University of Copenhagen


The Estimation of Bayesian Networks in the Presence of Exogenous Variables


The estimation of Bayesian networks given high-dimensional data sets, with more variables that there are observations, has been the focus of much recent research. While there are many methods available for the estimation of such structures, these methods typically assume independent and identically distributed samples. However, often the data available will have a more complex mean structure and additional components of variance. For example, in the estimation of a Bayesian network from gene expression data, the data set may contain exogenous variables thought to affect the expression levels of the genes of interest.

In this talk, I will be considering the case where the effect of such exogenous variables is not of primary interest, but secondary to the estimation of a Bayesian network given the data. After a brief review of Bayesian networks and the estimation thereof given high-dimensional data, I will discuss how exogenous variables may be incorporated into the model and why the incorporation of such variables in necessary.


Dr Jessica Kasza is currently a postdoc in the Department of Mathematical Sciences at the University of Copenhagen. She recently received her PhD from the University of Adelaide for a thesis entitled “Bayesian Networks for High-dimensional Data with Complex Mean Structure”. Her research interests focus on the theory of Bayesian networks and graphical models and the use of these structures in the modelling of genetic regulatory networks.

Date: Wednesday March 17, 2010


Details of AGM Elections and Agenda


Mr Paul Sutcliffe

Australian Bureau of Statistics


Australian Statistical Conference, Adelaide, July 2012


The fifteenth Australian Statistical Conference was last held in Adelaide in July 2000 at the Hilton. It is now official that the SA Branch of the Statistical Society will organise the conference in 2012 and I am sure that we will be able to deliver a great conference. In this presentation I will talk about the general themes that have emerged from reviewing evaluation reports and speaking to past organisers. In starting the planning process I have been assisted by Stuart McLeod from the Adelaide Convention Tourism Authority who has been invaluable in putting together suitable venue options. We have chosen July as the most suitable month to hold the conference and I will present three conference models for consideration of members.

Date: Wednesday November 25, 2009

SPECIAL EVENT – Student forum


Mr Nicholas Wilkey

PhD student, School of Politics, The University of Adelaide


Suicide attacks in Af-Pak


Suicide attacks are a huge ongoing threat in Afganistan and Pakistan. However surprisingly little research has been conducted which specifically examines this threat, as opposed to the considerable amount of work which deals with other aspects of the Afghan and Pakistani conflicts. My thesis research therefore examines the applicability of the major existing theories of suicide terrorism to the cases of Afghanistan and Pakistan in order to redress this situation. My talk will be drawn from research I am conducting for a chapter which deals with organizational theories of suicide terrorism. These theories attempt to explain the phenomenon in terms of the strategic and tactical benefits that suicide attacks provide those groups who employ them. I will present a brief histoty of suicide attacks and of the Afghan and Pakistani conflicts and then turn to an outline of the main theories relevant to this research. Finally, I will give an overview of the available relevant data and I will describe how I have been attempting to use it to test the competing theories.

Date: Wednesday October 28, 2009

SPECIAL EVENT – E. A. Cornish Memorial Lecture


Dr Louise Ryan

Cheif, CSIRO Mathematics, Informatics and Statistics


Data, data everywhere!


We live in data-rich times. Advances in information technology now allows us routine access to massive amounts of data through a variety of sources and relating to almost every aspect of daily life. This presentation will focus on the importance of modern statistical analysis and visualization tools to help process and turn these massive datasets into usable information, leading to new insights and, ideally, policy changes. Two examples will illustrate how routinely collected emergency room data can be used to inform decision makers about emerging diseases such as swine flu as well as understand socio-economic impact on heart disease.


Dr Louise Ryan is well known for her contributions to statistical methods for cancer and environmental health research. At CSIRO, she is currently leading a group of 150 people in mathematical and statistical research areas as diverse as financial risk, climate change and cell biology. This research is allowing CSIRO to better address national challenges.

Dr Louise Ryan attained a Bachelor of Arts in statistics and mathematics from Macquarie University Sydney, and a Doctorate of Philosophy in statistics from Harvard University, USA. Louise grew up in Australia but has spent almost thirty years in the USA, where she most recently held the post of Henry Pickering Walcott Professor and Chair of the Department of Biostatistics at Harvard University. She joined CSIRO in 2009.

Over Louise’s career she has been recognised with a number of professional awards and achievements including: Fellow of the American Statistical Association; the Spiegelman Award, conferred by the American Public Health Association; and the Distinguished Achievement Award of the Environmetrics Section of the American Statistical Association. Louise has also served as editor or associate editor for a number of statistical journals and President of the Eastern North American Region of International Biometric Society.

Other significant contributions Louise has made as a mentor to women and minority students have led to several awards which include the annual Mentors Award from Harvard School of Public Health and a Role Models award from Minority, Inc. Louise has been a passionate advocate for diversity in higher education and was the founding director of a program for the training of minority students at Harvard.

 Date: Wednesday September 16, 2009

SPECIAL EVENT – Joint meeting with the Australasian Epidemiological Association


Associate Professor Peter Baghurst

Women’s and Children’s Hospital and The University of Adelaide


Current issues in the monitoring of outcomes of care in Australian hospitals


The safety, quality and appropriateness of healthcare in Australia is now firmly on the political agenda, with Federal Health Ministers of both the former Liberal Government and the current Labour Government calling for greater accountability of hospitals through the public disclosure of key performance indicators. A major difficulty, however, is the dearth of relevant information required to generate meaningful comparisons – and a very weak understanding of issues like bias, confounding and hospital size, which makes the interpretation of ‘league’ tables difficult, – and very prone to mischief.

Death in hospital (or in the 30 days post discharge) is a very basic measure of the quality of healthcare – but the information on which to base adjustments for the many factors which impact on the risk of death is very limited. Graphical techniques such as funnel plots are useful for identifying hospitals whose performance deserves closer scrutiny either as opportunities for improvement – or as role models, but smaller hospitals often appear as the better performers, emphasising the need for more detailed information about risk factors. An application of risk-adjusted funnel plots to the obstetric outcome ‘Intact Lower Genital Tract’ will be used to demonstrate some of the difficulties of comparing hospital performances in a manner which will actually engage the clinicians.

Queensland, post Jayant Patel, leads the country in terms of applying statistical process control techniques to a wide variety of hospital outcomes – but the way in which alarm thresholds and Average Run Lengths are manipulated in order to force closer scrutiny of practice generates considerable workloads – often in response to false alarms. An example of how these methods might be used to influence obstetric practice will be provided.


Peter Baghurst is Head of the Public Health Research Unit, Children Youth and Women’s Health Service, Women’s and Children’s Hosptial – and Associate Professor in the Disciplines of Paediatrics and Public Health in the Faculty of Health Sciences, University of Adelaide.

Date: Wednesday July 22, 2009


Mr Michael Leonard

The University of Adelaide


Gaussian Random Fields for Space-Time Rainfall Modelling


The hydrologic response of urban catchments is sensitive to small scale space-time rainfall variations. A stochastic rainfall model used for design purposes must reproduce important statistics at these small scales. However, current models make simplifying assumptions about the way rain fields evolve and thus cannot be expected to reproduce important statistics over various space and time scales. In this talk, a new phenomenological heirarchical stochastic model is developed to robustly simulate rainfall fields consistent with 10-minute 1-km length pixel radar images. The heirarchical framework has three levels. The first level simulates a latent Gaussian random field conditioned on the previous time step, which is transformed to a rain field using a power transformation. A Toeplitz block ciculant technique is used to achieve fast and accurate simulations of large Gaussian random fields. In the second level, First-order autoregressive models are used to describe the within-storm variations of the level-one parameters that control the evolution of rain fields. The third level is designed for simulation of storm sequences, where the parameters of the level-two model are classified into different sets according to different storm types. Calibration is performed using a generalized method-of-moments approach. It is demonstrated that this two-level rainfall model produces realistic sequences of rain images which capture the physical hierarchical structure of clusters, patchiness of rain fields and the persistence exhibiting during storm development. Furthermore, a variety of important statistics are adequately reproduced at both 10-min and 1-hr time scales over space scales ranging from 1 km up to 32 km. Finally, application of this model to short-term rainfall forecasting in presented.


Michael Leonard completed his studies in Civil Engineering at The University of Adelaide in 2002. His PhD in hydrology stems from the need for stochastic models to assess the risks associated with flooding for engineering design. His interests are in spatiotemporal modelling, random fields and point processes and computer programming. As an example of the latter Michael has recently developed a library to allow graphic and packages in R to be accessed from a Fortran program ( and is currently teaching programming to Engineering students.

Date: Wednesday June 17, 2009


Dr Alun Pope

Rismark International


Statistical models for valuing residential properties on a large scale


Traditional methods for valuing residential property are expensive and the valuations they provide rapidly become out of date. Statistical models are much cheaper (once built) and can produce timely valuations for both individual properties and large portfolios. These models can be used for several purposes, including:

  • assisting in the lending decision;
  • regular valuation of portfolios of mortgage-backed securities;
  • construction of small area price indices; and
  • construction of a tradable daily price index.

This talk will cover statistical aspects of the design and implementation of models on an Australia-wide scale, with illustrations from the construction of the Rismark Automated Valuation Model, which is based on a combination of hedonic models.


Alun works for Rismark International, a Sydney-based company which carries out research and manages funds, specialising in residential real estate. Before that, he worked for St. George Bank and the Australian Prudential Regulation Authority. He was for many years an academic (University of Newcastle, UNSW), where his research interests included theoretical aspects of nonparametric regression and time series as well a applied work on problems relating to the assessment of risk on firing ranges. He was worked fro the ABS and the Department of Defence, and has been a partner in a small consulting company. His PhD was in pure mathematics at the University of London, but on returning to Australia immediately after that, he discovered the joy of statistics. He is an AStat and has been president of the NSW Branch of the SSAI.

Date: Wednesday May 20, 2009


Dr Helena Oakey

Senior Statistician
Discipline of Obstetrics and Gynaecology
School of Paediatrics and Reproductive Health
The University of Adelaide


Field trials, pedigrees and statistics: getting the mix right


This talk will give an overview of genetic variety testing from the breeders’ and statisticians’ points of view. The current designs used and approach to analysis of multi-environment trials will be discussed. The benefits of extending the analysis to include pedigree information will be presented. A practical example illustrating the methods will be shown.


Helena Oakey worked as a statistical consultant for the Faculty of Agriculture, Food and Wine (University of Adelaide) for six years before starting her PhD. Her PhD thesis looked at incorporating pedigree information into the analysis of agricultural genetic trials. She is currently working at ARCH (Australian Research Centre for Health of Women and Babies) overseeing the design and analysis of the clinical trials conducted by research staff.

2009 Annual General Meeting

The Annual General Meeting of the SA Branch was held on 18 March 2009 at which the 2008 Annual Report and Treasurer’s Report were presented and elections for branch council positions were held. The associated minutes and reports are available below.

SSAI SA Branch AGM Minutes 2009

SSAI SA Branch AGM Minutes 2008

SSAI SA Branch Financial Report 2008


Get the latest posts delivered to your mailbox:

Show Buttons
Hide Buttons