ewsletter June 2008 Number 123 N statistical society of australia incorporated Statistics at the 2020 Summit

Why I attended the 2020 summit development as a country interacting in Dr Harch was nominated by CSIRO’s the global environment. Chief Executive Officer to attend the 2020 The process undertaken to garner summit. Her international reputation as an ideas—where we were able to build on and environmental statistician, skill in leading adapt people’s ideas—highlighted that a multidisciplinary multi-organisational constructive dialogue enabling people to put projects and her more recent experience with forward ideas and having people actively leading international collaborations with listening ensures ideas can be generated China and the USA were seen as relevant through to a level of consensus. for making a significant contribution to Disagreement on ideas is also useful. ideas generated at the Summit. It enables discussion to be focused on My experience at the summit understanding others points of views— enabling a common understanding of I was part of the topic area “Australia’s issues. future security and prosperity in a rapidly Dr Bronwyn Harch trained as an changing region and world”. environmental statistician and currently Value of the summit The majority of people who participated has the role of Research and Business Already the Initial 2020 Summit in the 2020 Summit attended as Leader for CSIRO’s Environmental report is being used by federal and state “Australians”, but there were a few lobbyists Monitoring and Modelling research theme. The group’s research is focused governments in their deliberations about who attended as well. I found the 2020 on the application of mathematical and government priorities. I believe in the Summit as quite an inspirational experience statistical sciences for planners and policy- because I interacted with people from all process used to develop the ideas at the makers to make sound environmental Summit; a participatory process enabling walks of Australian life. Discussion with management decisions. participants—whether leaders in Australian consensus and areas of disagreement to be society or contributors to the fabric of openly discussed. Australia—were fascinating. for discussion in larger groups. The main During one of the group sessions, focus of the second day was to crystallise There was a sense of camaraderie in the Bronwyn shared with Kevin Rudd about shared development of ideas and a real and prioritise the generated ideas. While the idea of Australia adopting a new spirit of rolling up our sleeves together there was some contention about the approach using “smart power” to address and getting on with the task at hand. relative priorities of the ideas presented food, water and energy security issues in There was agreement on the majority of in the Initial 2020 Summit Report, a full collaboration with our neighbours. In this key challenges to Australia’s future security account of the ideas discussed will be case “smart power” is about information and prosperity, but there were differences collated into a Final 2020 Summit Report and knowledge platforms that enable the of opinion in relation to the solutions to in the coming month. prediction of future security issues and these challenges. allow governments at all levels to consider The process used to garner ideas had What I took from/learnt from scenario planning around this definition of us start on the first day with individual the summit security—which needs to be much broader reflection, then gathering as a small group The energy and passion of delegates than “defence”. “Smart power” is in contrast to share these individual ideas and then made it obvious that many Australians to “hard power” (use of military force) and synthesis of commonalities and differences care very much about Australia’s “soft power” (use of diplomacy).

In this issue… Editorial 2 Conferences/Workshops/Courses 8 President’s Corner 3 Workshop in Commemoration of Chris Heyde 10 Three Doors 4 In Memoriam 11 AGM Notice of SSAI and ASPAI 5 Branch Reports 15 2007 AusCan Scholar Report 6 Tea party held for t-test 20 Editorial

This issue appears to have a travel the t distribution to the world. Many theme underlying the features in PO Box 5111, people contributed to the success of the particular. At a local level, members are “t party”, including the Dean of the Braddon ACT 2612 continuing to travel and come together Faculty of Information Sciences and Phone (02) 6249 8266 for Branch meetings, and most branches Engineering, who provided financial Fax (02) 6249 6558 have reports in this issue. Bronwyn Harch support; the Faculty administration team, Email: [email protected] reports on her travel to Canberra for who decorated the space; two local Irish Society Web Page the 2020 summit, and Melissa Dobbie pubs, who provided the Guinness-themed reports on her travel to Canada as the http://www.statsoc.org.au decorations (the party was held just after AusCan scholar. It’s great to see female statisticians taking on high-profile roles St Patrick’s day so leftover decorations were easy to come by!); over 30 academics Editors both within in the Society and in the broader community. from all Faculties of the University of Canberra, all of whom used t tests in Alice Richardson, Faculty of ISE, I can’t finish this editorial without University of Canberra ACT 2601 mentioning the “t party for the t test”, their research; and invited guests from the Australian National University. We held at the University of Canberra in Michael Adena, Covance Pty Ltd March. This very successful and very hope that the report inspires you to come PO Box 5125, Braddon ACT 2612 enjoyable event was organised to up with your own event to celebrate the celebrate 100 years since the publication anniversary of this mainstay of statistical Correspondence of Student’s 1908 paper that introduced theory and practice.

Please direct all editorial correspondence to Alice Richardson. Email: [email protected]

Disclaimer Member News

The views of contributors to this Newsletter should not be attributed to Seventeen of Australia’s leading population dynamics of vertebrates; the Statistical Society of Australia, Inc. scientists were honoured on 19 March solvable lattice models; and mammalian by election to the Australian Academy of Subscriptions embryonic development. The new Science. Fellows, elected in 2008 from Australian The Newsletter of the Statistical Election to the Academy recognises a universities, CSIRO and medical research Society of Australia is supplied free to career that has significantly advanced, and institutions, include Professor Matthew all members of the society. Any others continues to advance, the world’s scientific Paul Wand FAA, distinguished for wishing to subscribe to the Newsletter knowledge. Scientific contributions of his work on nonparametric function may do so at an annual cost of the new Fellows cover a wide range of estimation. Congratulations Matt! A$30.00 for an issue of four numbers. specialities that include immune response to pathogens; physics of supernovae; Source: http://science.org.au/media/ Advertising photosynthesis; plasma physics; newfel2008.htm Advertising will be carried in the Newsletter on any matters which the Editors feel are of interest to the members of the Society. For details of advertising rates, etc. contact the SSAI Executive Officer at [email protected]

Printer

National Capital Printing 22 Pirie Street, Fyshwick ACT 2609

DEADLINE FOR NEXT ISSUE: 10 August 2008

 SSAI Newsletter – June 2008 President’s Corner

In this column I would like to reflect issues for Australian society. How would on the question: “Should the SSAI get we manage the reputational aspects of involved in commenting on significant these studies? The potential impact on societal questions where statistical our Society’s reputation is not to be analysis plays a substantial role?” underestimated—often the topics are This issue has been on the SSAI controversial and there may be widely agenda for some years without a lot differing views about them. How would so far being done to address it. The we finally agree that the conclusions SSAI Strategic Plan (currently being reached and the views expressed can be revised) lists as one of its strategies: put out under the SSAI banner? Any of “Convene specialist panels on significant our members who have given statistical community concerns that have a advice in courts of law will know that statistical aspect needing clarification quite often consensus on methods, and explanation.” I admit to having analysis and their interpretation is not given this issue low priority, partly reached by experts of equal standing. because I have not been able to work All of these questions, and others no out how the SSAI would implement this doubt, need to be addressed before the strategy. But, head in the sand or not, SSAI can put any strategy for this into as the professional society for statistical action. Right now we are hard pushed work in Australia we may need to give to get people to engage in running this attention. I have been asked to the SSAI on a voluntary basis I find it respond to two requests of this type over hard to imagine members engaging in Universities in Australia are offering a the past couple of months. The first was a disinterested way (rather than because 3-year sequence in Statistics. This issue for the SSAI to arrange an evaluation by they have a passion or view about highlights the need to lobby beyond Accredited Statisticians concerning the something) in these important issues Government and make sure the message conclusions drawn in a recent scientific without being renumerated in some way. is getting through at the University level paper that is claimed to have had a I would welcome member’s views on also. crucial influence on the recent Garnaut the above issues. In my capacity as a member, ex officio, Climate Change Review. The second was Members might not be aware that of the Australian Academy of Sciences concerned with evaluating the available an important role of the SSAI is to National Committee on Mathematical data and methodology used to arrive engage in various forms of lobbying. Sciences I recently contributed to the at the proper estimate of the gap in The key professional societies in the Submission to the National Innovation life expectancy between non-indigenous mathematical sciences have elected Nigel System Review prepared by Prof Hyam and indigenous Australians, a gap the Bean, Head of Applied Mathematics at Rubenstein who chairs that Committee. current Government is committed to the University of Adelaide to represent These are only some of the areas we closing. It was suggested that if this them on the FASTS Board. In his have been active in lobbying. were the UK then very likely the Royal inaugural submission Nigel raised several William Dunsmuir Statistical Society would be asked to set issues that FASTS will be encouraged Email:[email protected] up a working party to sort this out. to support. One of these was the fate The primary issue is: Should the SSAI of mathematics and statistics at the 1. http://www.amsi.org.au/pdfs/Questionnaire_ engage in these activities and is this University of Southern Queensland, summary.pdf engagement in the long-term interests an issue which has galvanized of the SSAI? If the answer is “yes” there mathematicians and statisticians around are several issues of implementation the world and one which has far wider that need to be considered if we are implications. In November last year I to effectively respond to such requests: wrote to the VC expressing concerns Who would do the work? Are only about the situation developing at USQ Accredited Statisticians the right people and informing him of our 2005 Review. to be involved? How would we select and Nigel will also alert FASTS about the prioritise the topics that the SSAI was fate of the new funding for mathematics prepared to evaluate? How would these and statistics teaching at Universities projects be funded? Funding might have that was included in the May 2007 Does SSAI have your to be provided or else the competing Federal budget. A survey conducted at current email address? demands on our members’ time will take the beginning of 2008 suggests that precedence; or else, members may only this funding is likely to be passed on to If you are not sure please let engage because they have a particular the mathematical sciences departments us know by email to: point of view or interest in the issue being in only a minority of universities1. addressed, hardly the right approach to This survey also suggests that we are [email protected] getting disinterested advice on important close to the point where only half the

SSAI Newsletter – June 2008  Three Doors with Borek Puza (Edition 14)

Welcome to the 14th edition of Bernoulli random variables, where the ..... is Euler’s constant. Thus n ~ exp(m Three Doors. Last time I presented The ith one has mean 1/i and variance 1/i - c), which gives results that are very Records Puzzle and am now pleased - 1/i^2. (Note that for a < b, whether close to those in the table. For example, to announce Conrad Burden as the the bth value is a record provides no if m = 6 then n ~ 226.51. current winner of the Three Doors information regarding whether the ath For further discussion on the topic Prize, a cheque for $60 from SSAI. value is a record. Therefore I(a) and I(b) of records, see Andeˇl (2001, Chapter 4 Several other persons also sent in the are independent. This fact can also be on pages 63-78). The author writes (on correct solution, which is given below, proved more formally.) Now, R is the page 65): “We read in [the table] that followed by a discussion and the next sum of the n indicators. Consequently, in a random series (such as maximal puzzle. m = 1 + 1/2 + ...+1/n and s^2 = m - (1 high of snow in winter) an 11-year-old + 1/2^2 + ... + 1/n^2). These formulae [boy] has seen in [his] life three records The Records Puzzle lead to the required table: on average; a gentleman 31 years old, Consider a sequence of n continuous iid j n m s four records; and a man 83 years old, random variables, indexed 1,...,n, and 1 1 1.00000 0.00000 five records. Isn’t this a key to the fact that in our youth winters were colder define the ith one to be a record if it is the 2 4 2.08333 0.81223 largest amongst those indexed 1,...,i. Let R with more snow, summers were warmer, 3 11 3.01988 1.20907 and, generally, everything was much be the total number of records amongst the better?” n random variables. Find the mean, m, 4 31 4.02725 1.55372 and standard deviation, s, of R. Hence, for 5 83 5.00207 1.83551 The Ten Digits Puzzle each j=1,2,...,10, find the smallest value of 6 227 6.00437 2.08898 In how many ways can the ten digits n such that m is at least j. Then, for each 7 616 7.00127 2.31473 such value of n, calculate the corresponding 0,...,9 be written in a row so that each exact values of m and s. Tabulate your 8 1674 8.00049 2.52114 digit other than the left-most is within results to 5 decimals and discuss briefly. 9 4550 9.00021 2.71210 one of some digit to the left of it? 10 12367 10.00004 2.89053 For a chance to win the next Three Solution Doors Prize, send your solution to [email protected]. Amongst the first i values, each one is Discussion equally likely to be the largest, and so the The harmonic series 1/1 + 1/2 + 1/3 References probability of the ith one being a record + ... sums to infinity. Thus m diverges is p(i) = 1/i. Consider the indicator without bound, albeit very slowly. For Andeˇl, J. (2001). Mathematics of I(i) for the ith value being a record. large n, we may use the approximation Chance. New York: Wiley. Then I(1),...,I(n) are independent m ~ log(n) + c, where c = 0.57721 56649

Australian Statistical Conference 2008

30th June to 3rd July 2008 – Hotel Sofitel, Melbourne Confirmed International Speakers

± Enrivonmental Statistics – Adrian Raftery, University of Washington, Seattle ± Statistical Genetics and Bioinformatics – Elizabeth Thompson, University of Washington, Seattle ± Medical Statistics – Simon Thompson, Cambridge MRC Unit; Michael Hernan, Harvard ± Stochastic Modelling and Statistics in the Social Sciences – Fiona Steele, University of Bristol ± Statistical Methodology – Alistair Young, Imperial College Conference website: http://www.asc2008.com.au

 SSAI Newsletter – June 2008 NOTICE

The Annual General Meetings of The Statistical Society Of Australia Inc and The Australian Statistical Publishing Association Inc. will be held on

Monday 30 June 2008 at 5.30pm at the Sofitel Hotel, Melbourne.

SSAI Annual General Meeting—Agenda ASPAI Annual General Meeting—Agenda

1. Apologies and Proxies 1. Apologies and Proxies Proxies must be given in writing as per form inserted Proxies must be given in writing as per form inserted in the June 2008 issue of SSAI Newsletter. Proxy forms with June 2008 issue of SSAI Newsletter. Proxy forms must be received by the SSAI Executive Officer for must be received by the SSAI Executive Officer for passing to the Secretary no later than 24 hours before passing to the Secretary no later than 24 hours before the time of the meeting. the time of the meeting.

2. Confirmation of the Minutes—Minutes of the meetings 2. Confirmation of the minutes—Minutes of the meetings as circulated as circulated

3. Matters arising 3. Matters arising

4. Reports 4. Presentation of the Annual Report by the Editor of the 4.1 President Australian and New Zealand Journal of Statistics 4.2 Treasurer 4.3 Branches 5. Treasurer’s Report 4.4 Sections 6. Appointment of signatories 5. Accreditation 5.1 Report from Accreditation Committee 7. Other business

6. Conferences 8. Time and place of next meeting. 6.1 ASC 2008 6.2 ASC 2010

7. Election of Section Chairs Nominations for Section Chairs should be received at the SSAI office no later than 23 June 2008. Nomination Forms have been inserted in each copy of the June issue of SSAI Newsletter. All nominations require a seconder and a statement from the nominee that she or he is prepared to stand.

8. Appointment of signatories

9. Other business

10. Time and place of next meeting.

SSAI Newsletter – June 2008  2007 AusCan Scholar Report

Between 14 January and 15 February The specific activities during my visit 2008, I visited Canada as the 2007 AusCan in each of these places varied and ranged Scholar, the first recipient from Australia from being pre-organised timetables and the second recipient of the scholarship of one-on-one conversations to more (following Dr Mu Zhu’s visit to Australia focused and deliberate discussions with in 2006). faculty members (and also with doctoral The broad aims of the scholarship and postdoctoral students at SFU). were to: Topics of conversation ranged from discussing my working life in CSIRO to 1. promote scientific interaction between details of my PhD research (on modelling the two statistical communities, and correlated zero-inflated count data) to 2. provide opportunity to enhance my Visiting the Department of Mathematics my current research and applications in professional development and interact and Statistics at McMaster University. spatial design for aquatic monitoring. with leading researchers in Canada. The latter area was summarised and the As such, I opted to focus my five-week two days visiting the University of Toronto focus of two seminars that I delivered visit on a few select places rather than try (UofT) in Toronto, and five days visiting on seven occasions during my visit; the to get around the whole country and meet the University of British Columbia details of these with copies of the slides as many Canadian statisticians as I could. (UBC) in Vancouver. I intended to visit are provided in Appendix 2. In particular I spent six days visiting Simon the University of Waterloo (UofW) for a The conversations I engaged in were Fraser University (SFU) in Vancouver, a day too but a winter snow storm in the always a two-way road; I was just as week visiting Dalhousie University (DalU) area closed the university on that day! A eager to hear about others’ research areas, in Halifax, three days visiting McMaster winter snow storm also cut short my visit working lives and statistical interactions, University and the National Water to the University of Toronto. The itinerary with the express purpose of enhancing Research Institute (NWRI) in Hamilton/ provided in Appendix 1 details the timings and helping to direct my own statistical Burlington,C080885_ANZJPH. one day visiting the University pdf Pageof these 1 visits 11/ whilst 4/ selected 08, photos 2:06 from PM research. There were two common threads of Western Ontario (UWO) in London, my trip are provided in Appendix 3. of statistical research that I identified from

AIHW Statistical Consultants - Statement of Requirement The Australian Institute of Health and Welfare plans to establish a panel of skilled statisticians for consultancy work. We are seeking statisticians with proven capability in applied statistics, who are able to assist with complex statistical problems and/or who have demonstrated ability to transfer knowledge to project staff.

Specifi cally, we are seeking individuals or a company with the ability to: • Interpreting and presenting research fi ndings to peers in the research community; • Develop and implement statistical methodology and/or adapt existing methods to meet • Participating in the development of new research projects; the AIHW needs. • Performing other duties as required. • Be part of a collaborative multidisciplinary research team and ensure skills transfer. • Write project reports in conjunction with other AIHW staff. Working Arrangements • Contribute to [preparing and writing] papers for publication in the scientifi c literature, with The Consultant(s) will on occasions be required to either: a focus on both statistical and applied journals and conferences. • work as integrated member of the AIHW team working at the Institute’s Fern Hill Park • Interpret and present research fi ndings to peers in the research community. • Participate in the development of new research. premises; or • provide advice and consultancy services ‘on line’ from their own accommodation. Position Description Remuneration Panel members will be involved in a range of services including: For consultancies that require an integrated team member, payment will be based on the • Providing a statistical consultancy and information service for AIHW staff; • Developing statistical methodology and/or adapting existing methods to meet the AIHW relevant salary rate for the expected duration of the project. research objectives; For ‘on-line’ consultancy services payment will be by an agreed hourly rate paid on receipt of • Implementing statistical methodology and working with AIHW staff to write project a monthly invoice. reports, ensuring the transfer of skills; • Either leading, or working as an integrated part of, a collaborative multidisciplinary For further information: Please contact Dr Phil Anderson, phone 02 6244 1125, or email research team; [email protected]

hmaC080885

 SSAI Newsletter – June 2008 my discussions across all locations I visited, namely in functional data analysis and in state space modelling (also known as hidden Markov Models or dynamic linear models). These were topics that I was keen to learn more about, especially in relation to applications of topics of relevance to my work in environmental monitoring. I met with Jim Ramsay (the Canadian father of Functional Data Analysis methods) briefly in Vancouver during my last week in Canada and attended a seminar he gave at SFU which described some of his recently published research on parameter cascading in high dimensional data. Other Canadian statisticians whom work in this area and that I had brief A beautiful day in London, Ontario, with John Braun, Duncan Murdoch, conversations with include Jiguo Cao and David Bellhouse and Wenqing He. Dave Campbell (both faculty members at SFU) and Fang Yao (faculty member also work in this area, namely Abdel El- • directly enhance my knowledge about at UofT). Shaarawi (NWRI/McMaster) and Grace cutting edge statistical research being Chiu (UoW). This provided opportunity undertaken in Canada; Rachel Altman (faculty member at to discuss some of my burning statistical SFU) helped explain the fundamental • meet and interact with selected questions on topics in this area but also to underlying theory of Hidden Markov Canadian statisticians, some who work establish stronger links with both of them Models, and Joanna Mills Flemming and in similar environmental areas, thereby for future potential collaborative research Mike Dowd (both faculty members of identifying potential future collaborative endeavours. DalU) explained how they were using researchers; and state space models to combine physical At SFU, where there is a strong • discuss my research in some detail, which graduate and postgraduate presence in the models with observed data in marine helped to clarify my own understandings Department of Statistics and Actuarial ecology applications. I also attended and exposed others to areas of research Science, I interacted with many of the a seminar presented by Mike at the and application that are practically doctoral and postdoctoral students and Bedford Institute of Oceanography in challenging but require statistical rigour. Halifax on data assimilation for ecosystem heard briefly about their research on a Overall, the trip was intellectually models, using state space models. Finally variety of topics including recurrent event Yiping Dou, who has almost completed data with medical applications, capture- stimulating and motivating but I admit 1 her PhD with Jim Zidek at UBC, ran recapture methods for salmon populations, that it was quite exhausting , even with through some R code she has written for functional data analysis approaches to such a seemingly relatively relaxed itinerary implementing dynamic linear models to modelling returned salmon abundance compared to Dr Zhu’s itinerary for his form spatial predictions (of ozone in her given river flow, and prediction of Australian visit. None-the-less, I believe example) at unmonitored locations from forest fires. This further broadened my I fulfilled the aims of the scholarship data collected from a sparse number of appreciation for research being undertaken and hope that the majority of Canadian monitoring sites. in environmental applications in Canada. statisticians I visited and spoke with gained Given my own research area and There were numerous positive mutual benefit from my visit. I thank both application focus is in designing outcomes of my visit to Canada as the of the societies with awarding me such a monitoring programs for assessing aquatic 2007 AusCan scholar. I regard the most unique and worthwhile opportunity. ecosystem health, I arranged to meet with beneficial of these as presenting me with One improvement for next time is to two specific Canadian statisticians who an opportunity to: advertise the scholarship and its aims more widely amongst the statistical communities. I found that there were very few Canadian statisticians who had heard about the program or knew why I was visiting. Consequently, I needed to continually explain what had brought me to Canada and the aims of my visit, to the point where I copied a paragraph down from the SSC website which I got my seminar introducers to read out before I spoke! Melissa Dobbie

1. possibly as a result, a “Vancouver” flu got the better of me on my final weekend in Canada but ended up just slowing me down in my last Chris Field treats me to my first whole lobster dinner! week in Canada.

SSAI Newsletter – June 2008  Society Secretaries Central Council: Conferences/Workshops/Courses President: Professor William Dunsmuir Secretary: Dr Doug Shaw International Symposium on Forecasting [email protected] 22-25 June 2008, Nice, France Canberra http://www.forecasters.org/isf/ President: Dr Glenys Bishop Secretary: Dr Ray Lindsay Australian Statistical Conference 2008 [email protected] 30 June-3 July 2008, Melbourne, VIC New South Wales http://www.asc2008.com.au President: Dr Eric Beh Secretary: Dr Boris Choy AMSI/SSAI ASC 2008 Satellite R workshop [email protected] 27-29 June 2008, Melbourne, VIC Queensland http://www.maths.anu.edu.au/~johnm/R-ASC2008.html President: Dr Miranda Mortlock Secretary: Helen Johnson ASC2008 Satellite Short-Course: A New Approach to Meta-Analysis [email protected] 4th July 2008, Melbourne, VIC South Australia www.statsoc.org.au President: Dr Andreas Kiermeier Acting Secretary: Paul Sutcliffe Bayesian Methods in Health Economics [email protected] 14-16 July 2008, Sydney, NSW Victoria http://www.arcs.com.au/documents/Bayesian_Syd_14Jul08.pdf President: Professor Mervyn Silvapulle Secretary: Lucy Busija 7th World Congress in Probability and Statistics [email protected] July 14-19, 2008, Singapore Western Australia http://www.ims.nus.edu.sg/Programs/wc2008/index.htm President: Marty Firth Secretary: Rebecca O’Leary ISBA 2008 Satellite Workshop: Bayesian Environmetrics [email protected] 17-18 July 2008, Brisbane, QLD Section Chairs www.isba2008.sci.qut.edu.au Bayesian Statistics International Society for Bayesian Analysis Conference Kerrie Mengersen 21-25 July 2008, Hamilton Island, QLD [email protected] http://www.isba2008.sci.qut.edu.au/ Biological Sciences Ari Verbyla Pre-Olympic Congress on Computer Science in Sport [email protected] 5-7 August 2008, Nanjing, China

Environmental Statistics http://www.olympiccongress.org Petra Kuhnert – [email protected] Australasian Conference on Mathematics and Computers in Sport Industrial Statistics 31 August - 3 September 2008, Coolangatta, QLD Ross McVinish – [email protected] http://www.anziam.org.au/MathSport/Conference.html Statistical Computing Kuldeep Kumar New Zealand Statistical Association Conference [email protected] 1-2 September 2008, Hamilton, New Zealand Statistical Education (co-chairs) The call for papers will be linked from http://nzsa.rsnz.org/NZSA2008/index.htm Michael Martin [email protected] NatStats08 Conference Peter Howley 19-21 November 2008, Melbourne VIC [email protected] http://www.nss.gov.au/natstats Surveys and Management Veronica Rodriguez International Conference on Applied Probability and Statistics (with [email protected] emphasis in Business and Industrial Statistics) Young Statisticians (co-chairs) 1-3 December 2008, Hanoi, Vietnam Richard Hutchinson http://www.action-m.com/CAPS2008/ [email protected] Pamela McCaskie Australasian GenStat Conference: [email protected] Biometrics in Primary Industries and the Environment Social Sciences 2-5 December 2008, Marylands Country House, Yarra Valley, VIC Michele Haynes – [email protected] www.dpi.vic.gov.au/genstat2008 Further contact details for Society Secretaries and Section Chairs Australian Statistical Conference 2010 can be obtained by contacting the 6-10 December 2010, Perth, WA Society on (02) 6249 8266

 SSAI Newsletter – June 2008 NatStats Conference08

The NatStats Conference08 will be held at the Crown Promenade, Melbourne from 19-21 November 2008. Preprogram registration for the conference is NOW OPEN! Register and pay before 20th June to receive a special price on your conference registration— even cheaper than early bird! a national perspective. The conference challenges in creating a culture of The NatStats08 web site, http://www. will be divided into three main themes: evidence-based decision making nss.gov.au/natstats, is your one stop shop Informing the Nation; Measuring the for all your NatStats08 needs. From the • Share your experiences and raise Progress of Society and Informing the web site you will be able to register for awareness of the significant issues Environment Debate. There will also be the conference, download our sponsorship within your field showcase sessions that will demonstrate and exhibition document as well be Information Platforms and State planning • Network with policy makers from informed of all the latest program and initiatives. across government, industry, academia social events developments. Participating in NatStats08 will and the community The conference program has been provide you with the opportunity to: If you would like to talk about the designed with policy and decision conference program please call Mark makers in mind. A range of speakers • Learn from experts and experience Lound on 02 6252 6325 or email him at covering overseas organisations, senior first hand the latest developments and staff from policy departments, as well the future directions of key national [email protected]. as representatives from academia, the statistics To register your interest in the private sector and the community are • Connect with the statistical conference please contact Conference being invited to help elevate thinking to community on the issues and Solutions on [email protected]

http://www.isba2008.sci.qut.edu.au/

ISBA 2008 Satellite Workshop Bayesian Environmetrics

17-18th July, 2008 Queensland University of Technology, Brisbane, Australia

Outline

This workshop is aimed at researchers and practitioners who are interested in Bayesian approaches to statistical modelling and analysis of environmental issues. The workshop will comprise four themes: (1) environmental modelling for air, land and water; (2) environmental risk and biosecurity; (3) climate change, and (4) new statistical and computing technologies.

Confirmed International Speakers

u Professor Robert Wolpert u Professor Jim Clark u Professor Katya Ickstadt Duke U., USA Duke U., USA U. Darmstadt, Germany u Chris Jewell u Dr Birgir Hrafnkelsson u Professor David Draper U. Lancaster, UK U. Iceland, Iceland U. California., USA

Register via ISBA2008 webpage: www.isba2008.sci.qut.edu.au

SSAI Newsletter – June 2008  Thinking Statistically Workshop in Elephants Go to School A UNIQUE TEXTBOOK Commemoration By A new way to learn Sarjinder Singh statistics using pictures, of Chris Heyde jokes, and tales. Reviews: Professor Chris Heyde passed away on 6 March Collins Carbno, Technometrics, A lot of learning with 2008. He had an exceptionally distinguished 2007, 49(4), 496. fun through 651 pages. Marcin Kozak, Statistics in career in probability and statistics. Transition,2006, 7(6), 1407-9. His research interests were broad, covering Good for all ages probability theory, applied probability and + Good for all libraries statistics, as well as the history of statistics. He + Good for all majors published over 200 papers and authored and + Good for all schools edited 12 books. + Good for you too Chris served the Australian Academy of PLEASE HAVE A LOOK Science, the Australian Foundation for Science, The Australian Mathematical Society, The Statistical Society of Australia, The International Statistical Institute and the Bernoulli Society by taking on significant roles in these organisations. He also acted as Editor or Associate Editor for Kendall/Hunt Publishing Company 4050 a number of journals, including the two Applied Westmark Drive, P.O. Box 1840 Dubuque Probability journals of which he was Editor-in- Iowa 52004-1840, U.S.A. Chief from 1990 to 2007. www.kendallhunt.com Chris was a Fellow of the Institute of Mathematical Statistics (1973), a Fellow of the Australian Academy of Science (1977), an Honorary Life Member of the Statistical Society of Australia (1981), and a Fellow of the Academy of the Social Sciences in Australia (2003). He became a Member of the Order of Australia (AM) for services to Mathematics (2003). He was MASA awarded a DSc honoris causa by the University of Sydney (1998) and received the Pitman Medal An International Journal (Statistical Society of Australia, 1988), as well as the Hannan and Lyle Medals (Australian Academy of Science, 1994 and 1995). Editor-in-Chief: Sarjinder Singh The Australian National University and the Co-Editor-in-Chief: Marcin Kozak University of Canberra are organising a workshop to commemorate his contributions to Statistics, Managing Editor: Stephen Horn to Academia and the statistical community in Treasurer: Sylvia R. Valdes Australia and overseas. The workshop will be held on 29 August 2008, at the Council Room, Building 1 of the University of Canberra. At Welcomes to: this stage, speakers and participants include Ejaz Ahmed (Windsor University, Canada), Bob Sampling, Econometrics, Bayesian Anderssen (CSIRO), Daryl Daley (ANU), Joe Statistics, Time Series, Design of Gani (ANU), David Heath (ANU), Shuangzhe Liu (UC), Ross Maller (ANU), Richard Experiments, Multivariate Analysis Morton (CSIRO), David Pederson (UC), Alice Richardson (UC), Peter Taylor (UC/AMT) and Alan Welsh (ANU). www.iospress.nl All colleagues and friends are welcome to attend. Further information is available from Alan Welsh ([email protected]) or Shuangzhe Liu ([email protected]).

10 SSAI Newsletter – June 2008 In Memoriam

Christopher Charles Heyde, in Mathematical Statistics in the first 20 April 1939 – 6 March 2008 full year, 1960, of the existence of the It is with great sadness that we report Department of Mathematical Statistics at on the passing of Chris Heyde. Chris, as the University of Sydney, in which Oliver he was known to us, was a giant figure Lancaster was the foundation Professor, in Australian Probability and Statistics graduating with a University Medal in during a period of almost 40 years. His 1961. He continued his studies at Sydney, active research career, stretching right up and received his MSc in 1962 for a thesis to the last weeks of his life, spanned an era on the “Theory of characteristic functions when a number of other major Australian and the classical moment problem”. His note published in 1963 in which he figures strode both the Australian and showed that the lognormal distribution the world stages. Some of these were his is not determined by its moments, is mentors, some were colleagues, some were from this period. It became a classic, and his students, many were his friends, but achieved the accolade of being cited in only a few had the impact, overall, that William Feller’s classic Introduction to Chris had. His influence went well beyond Probability Theory and Its Applications, the confines of narrow academic research, Vol.2. as he lobbied hard in corridors of power for recognition of the applicability of In 1961, Chris began work on a PhD Probability and Statistics in contributing thesis in Pat Moran’s Department of to solutions of the large questions facing Statistics at the ANU. The PhD was our times. He took a serious interest in awarded in 1965, and later that year he the development of our subject, both married Elizabeth (Beth) James, whom in Australia, and internationally. His he had met at the ANU, while both publications list contains invited articles awarded our ultimate honour, the Pitman were engaged on their PhD studies. which attest to his ongoing concern for Medal, in 1988. The mutually supportive marriage was the public perception of, and the future Active also in many other arenas, Chris remarkably happy. They had two boys, of, mathematical and statistical science, was elected Fellow of the Australian Neil, born in 1967, and Eric, born in presented from his authoritatively Academy of Science in 1977, and was 1969. At the time of Chris’s death, he perceptive standpoint. a Member of Council of the Australian and Beth were the proud grandparents of four grandchildren. While representing Australian Mathematical Society in 1980-83, and its interests in a substantial way overseas, Vice-President in 1981. The University In September 1964, Chris joined of Sydney conferred a DSc, Honoris via membership of major statistical Joe Gani, then also a member of Pat Causa, upon him in 1998, and he became groups such as the ISI and IMS, and Moran’s Department, in moving to the a Member of the Order of Australia (AM) the holding of a position at Columbia Department of Statistics at Michigan in 2003, for his services to Mathematics, University joint with his appointment at State University. Joe left there to take up more particularly to Statistics and the Chair of Probability and Statistics at the ANU, together with regular invited Applied Probability. In recognition of his the University of Sheffield, UK, at the appearances at a large number of the contribution to the social sciences he was end of 1965, and Chris followed him. He most prominent conferences, Chris did elected to Fellowship of the Academy of was soon promoted to Special Lecturer not neglect his support of Probability the Social Sciences in Australia in 2003. in charge of the Statistical Laboratory and Statistics at home. He always had a He was awarded the Hannan Medal at the University of Manchester in 1967, strong commitment to the SSA. Apart of the Australian Academy of Science when the Manchester-Sheffield School from early involvement as a local branch in 1994, and its Ranken Lyle Medal of Probability and Statistics was formed. member, he was Editor of the Australian in 1995. Overseas, Chris’s abilities were Journal of Statistics from 1973-78, and recognized early. He became a Fellow of Chris returned to Australia in President of the Canberra branch from the Institute of Mathematical Statistics September 1968, where he took up a 1977-78. In 1979 he was elected President in 1973. Later, he delivered the Fisher Readership in Ted Hannan’s Department of the Society, a position he held till 1981. Lecture at an Invited Paper Meeting at of Statistics at the ANU. He had by then He was a member of Central Council the 47th Session of the ISI in Paris. produced some 30 papers, a dominant theme of which was the refinement of from 1973-86, and the Society’s Federal Chris was a Sydney-sider, schooled classical limit theory involving large and President in 1985-86. Always active in at Barker College, Hornsby, where he small deviations, rates of convergence committee work, he chaired the SSAI excelled at sports, particularly swimming. Statistics Policy Committee from 1979- He was Dux of the School in 1956. A and domains of attraction, while 1984, and was SSAI Representative on damaged Achilles tendon on the football displaying a breadth of interest in the the National Committee for Mathematics field and a gifted mathematics teacher led contemporary issues in probability. The from 1981-1996. He was elected an SSAI him to redirect his energy to academic appointment stimulated Chris’s interests Honorary Life Member in 1981 and pursuits. He completed his Honours continued page 12

SSAI Newsletter – June 2008 11 In Memoriam from page 11 Society, responsible for probability and human heart within, and a strong ironic in new directions, notably the theory statistics, from 1972-74, and of Stochastic sense of humour, and the advice and of branching processes and statistical Processes and its Applications from 1972- encouragement he delivered impartially inference for them, and population 82 (he was its Editor from 1983-89). He stemmed from a genuine concern for the genetics models related to them. Several was Editor of the Australian Journal of well being of his students, his colleagues, joint papers with Eugene Seneta appeared Statistics from 1973-78, an Associate the authors of the myriad of papers he on these topics, and a fundamental Editor of the Annals of Probability, from handled in his working life, and all that one with Ted Hannan on time series 1974-81, and of Mathematics of Operations he came in contact with. analysis. In this work, a principal focus Research, from 1976-90. He was one Chris was diagnosed with leukaemia by Chris was the martingale concept. of the Editors of the Springer series of 11 years before his death, and underwent He was to become widely known for books in Probability and its Applications periods of treatment followed by lengthy work on the theory and application from 1985. His high standards, efficiency periods of remission. He completed his of martingale methods, not least in and integrity as editor, author and normal activities at Columbia University estimation for stochastic processes. In coauthor were always greatly respected by in the Fall of 2007, but early in 2008 1973 he was awarded a DSc by the ANU. his professional colleagues. metastatic melanoma was diagnosed in Canberra. In an email message dated In January 1975, Chris joined the Chris’s research demonstrated great January 20, 2008, to one of the authors of Canberra-based Division of Mathematics depth and originality in his broadly the present article he wrote: and Statistics, of the CSIRO, again chosen research areas, but beyond these he nurtured also a sustaining interest in the following in the footsteps of Joe Gani. Whatever happens, I certainly feel that I history of Probability and Statistics which On Joe’s departure as Chief of Division, have had a fortunate life. I will be happy to fostered an awareness of the important Chris took over as Acting Chief in 1981, have more, … but if not, I have had a good foundational elements of our subject. until September 1983 when he was innings and can go in peace. He authored and edited twelve books, appointed Professor and Chairman of the Chris Heyde was one of the greats of three of which are historical, and six of Department of Statistics at the University probability theory of our time and his which are edited collections of papers for of Melbourne. He was instrumental in contribution was immense. His influence special issues of journals, Festschrifts, or creating the Statistical Computing Centre will remain. overviews of particular topics. Two others there, and, in 1985, a Key Centre for represent major original contributions He will be painfully missed, by his Statistical Science, of which he became colleagues, his wife and family, his many to probability and statistics: Martingale the Foundation Director. This was a friends, and the statistical community Limit Theory and its Applications (1980) joint enterprise of LaTrobe, Monash and worldwide. with Peter Hall, and the later Quasi- Melbourne Universities, and the Royal likelihood and its Applications (1997). Eugene Seneta and Ross Maller Melbourne Institute of Technology. A more extensive appreciation of In May 1986 Chris returned to the References Chris’s life and works than the present ANU, and was Professor and Head of the one, and of the honours accorded to him, Gani, J. and Seneta, E. (Eds.) Department of Statistics in the Institute written by Joe Gani and Eugene Seneta, (2004) Stochastic Methods and Their of Advanced Studies from July 1986 to appeared as an introduction to Gani and Applications: Papers in Honour of Chris December 1988. From 1989-1992 Seneta (2004). This was a special issue of Heyde, J. Appl. Prob., Special Vol. 41A. he was Foundation Dean of the ANU the Journal of Applied Probability, which (Introduction by Editors: pp.vii-x.) School of Mathematical Sciences (now contained a collection of articles by his Glasserman, P. and Kou, S. (2006) the Mathematical Sciences Institute). colleagues and friends. It includes a list of A Conversation with Chris Heyde. Since 1993, while continuing at ANU, he his publications up till 2004; that list will Statistical Science, 21, 286--298. was also a Professor in the Department be updated in a collection of his articles, of Statistics at Columbia University, New accompanied by commentaries by a York, and Director of their Center for number of colleagues, in an IMS volume Daniel Dice Schubert Applied Probability. to appear next year. A published interview Many in the statistical community, One of Joe Gani’s major achievements of Chris appeared in Glasserman and Kou particularly the younger members here was the setting up of the Applied (2006). There will be an obituary by Gani in Perth, knew, or knew of, Daniel, and Probability Trust, and the publication of and Seneta, to appear in of the Journal of were shocked by his recent death which its two world class journals. Chris had a Applied Probability 2008, extending the happened in a motorcycle accident close association with these from their 2004 introduction mentioned above. last November. Daniel did his PhD inception, and became Editor-in-Chief of When first met, Chris could be a at Murdoch University completing the Journal of Applied Probability and of the somewhat austere and even forbidding successfully in 2006, and, at the time of his Advances in Applied Probability from 1990- figure. But no first impression could passing, was employed on a Postdoctoral 2007, jointly with Soren Asmussen from be further from the truth. His outward Fellowship at CSIRO Division of May, 2005. Other editorial contributions reserve, and the rigour he applied to Mathematics and Information Sciences included positions as Associate Editor his administrative duties as much as in Floreat. As the person who supervised of the Journal of the Australian Math. to his mathematics, concealed a very his PhD I consulted with Harri Kiiveri,

12 SSAI Newsletter – June 2008 In Memoriam his supervisor at CSIRO, and we thought was an excellent of games that could never happen in real we would put together some words to fillip to writing life. For example, he ran a series of test inform our readers about this unique his PhD. Daniel’s matches between Bradman’s team of the character. PhD was titled: 1940’s and Ricky Ponting’s team of 2006. From Brenton Clarke: Daniel came to “A Multivariate (Bradman’s team consistently won the Murdoch University after completing his Adaptive Trimmed series). He didn’t stop at this however. He BSc Hons 1st Class, which he completed L i k e l i h o o d would make up test cricket teams from Algorithm” externally to Murdoch. Initially he resided soccer players so for example Manchester in Sydney while doing his ordinary degree And a digital United would play Arsenal in a test Daniel Dice and he spent time in Canberra at the thesis is available at cricket series. He did this with classical Schubert holding Murdoch University composers also and simulated test cricket Bureau of Statistics while completing his “dice”. This photo Library http:// matches between Austrian composers honours year (again externally). Daniel is courtesy of his wwwlib.murdoch. was very diligent and asked questions partner Josie and German composers amongst others. and it was a tribute to him that the Hawkins. edu.au/adt/browse/ Its kind of interesting thinking about his Murdoch Mathematics and Statistics search/?k=Schubert score sheets with things like Beethoven Secretary at the time had a special button From Harri Kiiveri: Two years ago caught Mozart bowled Wagner for 67. I on her fax machine so that staff could Daniel Applied for and was awarded a don’t know exactly how he’d figure out if fax him back answers to his many and Post Doctoral Fellowship at the CSIRO Beethoven was a batsmen or bowler or wide ranging queries. Daniel was inspired Division of Mathematics and Information what his stats were, but it had something by the mathematics of robust statistics, Sciences in Floreat. I became Daniel’s to do with the impact the music had on though he was no slouch when it came to supervisor and mentor. Daniel getting his hands dirty with computing When Daniel started work, the first Looking back on my experiences with and analysis of data. After completing thing I noticed was that the quiet and Daniel it seemed that there had been a his honours, he took up the offer of peaceful life in my office had changed guiding hand directing Daniel. When I a Murdoch scholarship and decided to significantly. Daniel become a regular think of the last few years so many things move to Perth. visitor to my room for discussions, not began to fall into place like clockwork Daniel soon made an impression on only with Maths questions about projects • He achieved his goal to study the lives of people in our department we were working on, but on all manner mathematics—culminating with his and, indeed, many around campus. He of things ranging from current affairs, PhD religion, politics and the meaning of life. had fixed his mind on gaining a PhD. • He found a promising career path that Through these talks it became clear that Daniel was quite different, which we all really interested him and enabled him we shared a lot of similar (sometimes soon realized, but in a nice and caring to continue studying mathematics way. He revered female mathematicians, radical) views about things. Daniel liked • His finances improved had studied theology for the priesthood, to talk and very little was off limits to but had gone away from that, he loved talk about. Daniel had certainly been • He started to look after himself better games of chance, and spent much of his through hard times, harder than most, yet (improving on his Spartan student spare time designing games to play on he didn’t allow that to limit him. I don’t diet) the computer (hence the middle name exactly remember how Daniel’s interest in • He got his own transport mathematics began, however some time and the photograph). It became apparent • He began to expand his interests and ago he decided that he wanted to study that he loved cricket, soccer(Manchester social circle beyond work United), and most of all classical music, mathematics and with single minded • He met his life’s partner being one of the first in our department purpose he did whatever it took to do to purchase an Ipod, on which he stored that. This is what finally led him to Perth Unfortunately this progression was to study at Murdoch University where he 500 classical pieces. tragically cut short by an accident which was awarded a PhD in Statistics. was no fault of his own. It has been Having succeeded making the PhD difficult to come to terms with all this… project ‘work’ he gave an excellent talk Daniel had wide ranging interests at the Young Statisticians workshop at including Cricket (he used one of the As time passes we are left with Thompkins Park in 2005. We submitted algorithms he developed in his PhD memories…memories of Daniel’s a paper to ANZJS. This was a turning thesis to prove that Bradman was an enthusiastic work on real life problems point for him to write his thesis. As one outlier), soccer, classical music—in which required statistics… with seeing can expect referees comments are often particular Wagner’s operas, mathematics, his gradual “conversion” to Bayesian insightful, but must be answered carefully and more recently motor bikes. statistics…with the funny stories that and with precision. Sitting with Daniel Being a creative person Daniel had an seemed to accompany most things that he he soon learnt the craft of not only unusual way of combining his interests. did…with the energy and enthusiasm he writing a good paper, but finishing it off He would run computer simulations of put into his life…with his “what you see in a way that would please the editorship test cricket matches (in fact test series) is what you get” manner, and especially and the reader. The paper appeared in using the published statistics of all the with memories of his unique ability to the 2006 September issue of ANZJS, and players. He would simulate the outcomes create connections with other people.

SSAI Newsletter – June 2008 13 Looking for a job?

For a listing of current statistical vacancies in Australia and New Zealand visit:

http://www.statsci.org/jobs

Do you have a job to advertise on the website?

Email a position description to [email protected]. Listing is free!

This service is proudly brought to you by:

14 SSAI Newsletter – June 2008 Branch Reports

Queensland Branch News Time Zero-Inflated Count Models of design for generalized linear models Harbor Seals. (GLMs) has become a topical and integral February Meeting 3-Apr Following the AGM: Professor part of experimental design. This has Annette Dobson, School of Population predominately been a consequence of both On 26th February Professor Murray Health, University of Queensland spoke on their broad application and to the ever Aitkin and Dr Irit Aitkin from the Missing Data in Longitudinal Studies. growing computational power of modern University of Melbourne spoke at day computers. The problem is one of 8-May Dr Toshi Ohnishi of the Institute Queensland University of Technology. The choosing a design such that parameters can talk was titled ‘Multilevel analysis of item of Statistical Mathematics, Tokyo, spoke on Estimating a common slope (of multiple be estimated with high accuracy. GLMs are response models in clustered and stratified generally nonlinear models (nonlinear with survey designs for the US National strata) in the Tweedie generalized linear model using a conjugate prior. respect to parameters). For such situations, Assessment of Educational Progress’. it is necessary to have preliminary estimates 5-Jun Dr Dr Yougan Wang of CSIRO The National Assessment of Educational of parameters to form designs. This is Mathematical and Information Sciences, Progress (NAEP) is a series of US national indeed the main difference between design Indooroopilly, spoke on Smoothing for tests of performance in primary and high for linear and nonlinear models; the former estimating asymptotic covariance matrices schools across a range of subjects. Results not requiring information about the with applications in quantile regression and of the tests are published by the National unknown model parameters. These design rank regression with censored survival data Center for Education Statistics, in terms basics are initially revised as an introduction of differences in performance between 3-Jul UQ Professors Anthony Davison into some recent developments of design ethnic groups, national regions, and other and Louise Ryan presented talks. Professor criteria for GLMs. Davison spoke on Galaxies, ticks, neurons, important reporting groups. The primary A new criterion termed P-optimality for and stock market crashes: Hard times for school mathematics tests are composed GLMs which maximizes the probability of the Poisson process. Professor Ryan spoke of items which are analysed using item observing an outcome or set of outcomes on Optimal Design for Epidemiological response (IRT) models which regress is proposed. The development of this Studies. the binary response on a latent ability criterion was motivated by experiments variable, on which the reporting group 7-Aug QUT James Enoch, Academic where designs for the efficient estimation differences are based. The survey design Program Manager, SAS Institute, Australia of parameters produce low or modest of many of the NAEP tests is multi-stage and New Zealand spoke on Bridging the probabilities of observing successful with both clustering and stratification. This gap between Academia and Business outcomes. In an applied setting, despite complicates considerably the analysis of 20-Sep QUT Professor Doug Altman being able to efficiently estimate the test data. In this talk the multilevel presented a talk in Sydney which was parameters, ethical issues may be raised model-based analysis of one scale from viewed by Queensland members a via when conducting an experiment where only the 1986 mathematics test for children video linkup. Professor Altman’s talk was small probabilities of successful outcomes in Grade 3, age 9 was described. The full titled Selective non-reporting of findings of are expected. P-optimality combined with analysis required a four-level binary logistic randomised trials and impact on systematic parameter estimation provides a technique item response model. Current analysis reviews. for maximizing successful outcomes and methods for the NAEP surveys and the 16-Oct UQ Dr Timothy Robinson of minimizing uncertainty about parameter improvements which result from the full the University of Wyoming presented a estimates. For the criteria developed, a model analysis were discussed. talk titled “Bayesian Analysis of Split-Plot general equivalence theorem was shown to Experiments with Non-Normal Responses apply which provides theoretical support April Meeting for Evaluating Non-Standard Performance to the methodology and ideas presented. The AGM was held before the talk. Criteria”. Applications to design problems found in Dr Ross Darnell our outgoing President Nov-07 QUT Two day Workshop on the literature will be explored to show was thanked for his leadership in a very the R System on 26-27 November 2007 at the benefits of implementing this new successful 2007 program (see below). Dr the Queensland University of Technology. methodology. Miranda Mortlock takes over as President. Ten participants took part in the workshop Earlier this year, James submitted his The outgoing committee was thanked for which was delivered by John Maindonald, Ph.D thesis after 3 years of research at the all the hard work in the previous year. Many ANU. University of Queensland (UQ). During members are continuing in their role and in Dec-07 QUT Professor Helen this time, he was fortunate enough to have particular Dr Helen Johnson as secretary MacGillivray gave a vigorating talk the opportunity to study at the University and Dr James McBroom as Treasurer. The titled “Developing probabilistic statistical of Southampton and the University of full committee will be introduced in the thinking”. The meeting was held on Otago. James also gained experience in next newsletter. Tuesday the 11th December at QUT’s the application of current statistical The program from 2007 is summarised Gardens Point Campus at 5pm. methodology within the pharmaceutical here: James McGree from the University of industry as a student intern at Novartis 6-Feb Jay M. Ver Hoef, NOAA National Queensland spoke on ‘Probability-Based Pharma, Switzerland. James currently has a Marine Mammal Laboratory, Alaska gave Optimal Design’ on 15th April at the position at UQ as an associate lecturer. a very interesting presentation on Space- University of Queensland. In recent years, Miranda Mortlock

SSAI Newsletter – June 2008 15 Branch Reports Continued

Canberra Branch News Operating Characteristic curve which epidemiological data can also be required. compares the rate of false positive vs (1 - rate Michael explained how the statistical work Beyond the hype: data mining of false negatives) for the binary models. For he did provided support for the economic models used to evaluate the cost-benefit of in practice interval variables, they used the generalised c- statistic, which is the interval equivalent new drugs in terms of both life-years and The talk for the March meeting of the of the Receiver Operating Characteristic quality-adjusted life-years. In an aside he Canberra Branch of the SSAI was delivered, curve for binary variables. These tools were remarked that in pharmaceutical evaluation, following its Annual General Meeting, by used for choosing between models based saving 1 life-year was likely to be regarded our branch secretary, Ray Lindsay. Ray’s on training data at time of model building, as cost-effective up to about $40 000, talk was about his work as a senior data and evaluating over time based on actual whereas in road safety improvements the miner in the Australian Taxation Office, lodgments, subsequently. upper limit was closer to $500,000. and had an interesting title - “Beyond the In his specific example he discussed 2 hype: data mining in practice”. SAS/Enterprise Miner 5.1 was used for the two stage models. Models available studies of a new protease inhibitor which He opened the talk by referring to a July included regression, neural network, was expected to lower the HIV viral 2006 article from the Australian Financial Classification And Regression Trees load compared with control treatments. Review with headline “ATO data mining [CART] and others. Regression and The patients in this study were all ‘highly unearths treasure”. The article reported that “neural nets” do not cope with missing treatment experienced’ patients for whom the ATO reaped a $118 million dividend data. While CART coped better with other drugs had failed. These studies show from its first attempt at using data mining missing values, it only predicted a small a significantly higher proportion of patients to score taxpayers who are late filing returns. number of distinct values, and was not with a continuing response at 48 weeks Ray retold this story from a technical useful for discriminating between a large with the new drug compared with the perspective, highlighting the technical and number of units. However combining the control. However most of this difference statistical obstacles of the analysis, which component models in two stages increased seemed to be apparent at by week 8, with “made the journey very interesting”. the number of unique predictions beyond the time to treatment failure curves being The problem he tackled was to predict that expected. The team experimented with roughly parallel after then. revenue that would arise if a large number using ensemble models and these improved In discussing what this means for of overdue income tax returns (or business the percentage of unique values further. patients, and how much the community activity statements) were lodged by taxpayers Ray noted that it helps to break up a should pay for this drug, Michael said and to use the predicted revenue as scores problem to smaller parts, build relatively the standard economic approach was to for prioritising action. Many apparently simple models and combine them in the calculate the difference in life-years gained, overdue returns are in fact not required to right way. He concluded by quoting George which gave a cost per life-year. However, be lodged and the client has simply failed E. Box: “All models are wrong, but some as there were only 51 deaths (out of 1488 to notify the Tax Office. Separating the patients) during the study, a modeling models are useful.” wheat from the chaff would lead to a better approach is necessary. He analysed the outcome from revenue collection activities. Frank Yu trial data to estimate parameters for a Training data were used to fit the model Markov model with 12 ‘alive’ states and Developing treatment outcome models from lodgments that have been made and a death state. By modeling the transition the model was then used to derive score data from clinical trials data probabilities between these states in each from those overdue. There was no shortage At the Canberra Branch meeting on 8 week interval, using logistic regression, of data. Data items included previous 29 April, Michael Adena, Statistical Michael was able to show that treatment lodgment behaviour, known income and Consultant with Covance gave a talk, differences in transitions were in the first credits, demographics, whether a tax agent “Developing treatment outcome models 8 week period, with little or no treatment was used, industry codes (for business), and from clinical trials data”. Michael spoke differences after then. The study data so on. There were some 400 attributes and both about the general issue of statistical confirmed that the death rate was strongly each model used about 10 to 20 attributes. support for Pharmaceutical Benefits dependent on the current CD4+ count. Many data items were missing, and not Advisory Committee submissions and Michael then used an example from at random. Other interesting challenges about the specific issue of evaluating a oncology to motivate the use of Poisson included plenty of zeroes and long tailed particular drug for treating HIV. regression as an alternative to Cox distributions. Ray’s team used a two stage The PBS, established in 1948, aims regression for the analysis of survival. model, firstly to model the sign and then to give the Australian community access Whereas Cox regression does not give an the magnitude of the predicted values. All to safe, efficacious and cost-effective explicit baseline hazard, Poisson regression financial variables were log transformed medicines. Guidelines for getting a drug does. Poisson regressions produce relative with those that can be both positive and adopted on the PBS are seen as amongst hazard estimates close to those from Cox negative becoming two transformed the most rigorous in the world. Submissions regression, are simple to fit, and support a variables. require a detailed submission, following rich modeling framework. Given the focus was to derive priority new guidelines introduced in 2007. Reports In conclusion, Michael described some scores, normal metrics for model fitting typically include meta-analyses of previous of the fun aspects of his work, especially were less relevant than measuring a correct studies, extracting salient information. Re- the intellectual challenge, variety of work, rank order. Ray’s team used the Receiver analysis of clinical trial data and analysis of modeling, working with current medical

16 SSAI Newsletter – June 2008 Branch Reports Continued

issues and being involved in a world leading President’s two year term, the previous The coming few months also proves to area. His work requires integrating large President will be the Vice President, while be a very active period for the Branch. Our amounts of information from different in the second year a new Vice President monthly meetings will continue and, as sources, usually in a short time frame. will be appointed who will fill the role always, you will be kept informed of their Ray Lindsay of President at the end of the current details. President’s term. Phew. The change in Last July, the Branch hosted the workshop New South Wales the rules is now available on the Branch’s given by Kerrie Mengerson, “Workshop on website. After the AGM, Caro Badcock Branch News Bayesian Statistical Methodology”, which stepped down as Branch President to take saw 59 people attend. We are again going to up the newly formed Vice President role host another Bayesian flavoured workshop The past six months have been a and I was handed the reins with Boris from the 14-16 of July in collaboration busy period for the NSW Branch. On Choy (UTS) as the Branch Secretary. I with ARCS Australia and the APBG. This the 29th of November we had the 8th therefore had the pleasure of presenting workshop is aimed at health economists Annual J. B. Douglas Awards and Dinner the annual Lancaster Lecture where I and statisticians with the third day focusing at the Macquarie Graduate School of attempted to persuade those listening how more on advanced topics for the statistician. Management Conference Centre. These one can use various categorical analytic Professor John Stevens, from Sheffield awards are competitive and pit students tools to study yeast cells. I considered University UK, will be the facilitator and enrolled in research masters or PhD the use correspondence analysis, non- more details can be found at the Branch’s statistics degrees from universities across symmetrical correspondence analysis (there website. Also, the week after the ASC the state against one another. On the is a difference) and measures of association we are going to hold a joint session with night we had six excellent student speakers such as the Marcotorchino index and Grey- APBG and Pfizer again. Simon Thompson, and the Peter Wright Prize was jointly Williams index to help understand the Director of the MRC Biostatistics Unit in awarded to Stephanie de Silva (University factors that influence the ageing of a yeast the UK and one of the plenary speakers at of Sydney) and Dorothy Wong (Macquarie cell. I did make a disclaimer highlighting ASC, has agreed be out guest speaker. University). I would like to thank Caro that I’m not a molecular biologist, but You may recall that last July the Badcock (previous NSW President), my colleagues (Dr Vince Higgins, and Branch held a workshop “Integrating William Dunsmuir (UNSW and current our PhD student, Lloyd Mirto) are. The statistical ideas into mathematics” which Society President) and Jo Leadbetter (Eli talk discussed the findings of a study provided an opportunity for mathematics Lilly Aust PL) for being part of the undertaken to determine factors (including teachers, curriculum developers, university judging panel. Prior to the dinner William various oxidising agents) that inhibit a yeast statisticians and other representatives from Dunsmuir, our guest speaker, discussed cells ability to respire. the secondary schools from all around “Quasi-Monte Carlo in Statistical Model In April we had the pleasure of Professor NSW to discuss the proposed changes to Estimation” which aims to deal with David Hensher, Director of the Institute the NSW senior mathematics curricula. To problems that hinder the calculation of of Transport and Logistics Studies, follow on from the success of this workshop high-dimensional intractable integrals and University of Sydney, give a talk titled we are in the process of organising a second instead. He gave a general discussion of “How do respondents process stated choice education workshop to be held at end of the QMC methods and how they can be experiments? – Attribute consideration July, or early August (very soon). Helen used in statistical modelling, especially under varying information load”. David MacGillivray, a Carrick Award winner, has generalised linear models and time series spoke about the popularity of state choice agreed to be a guest speaker. models with latent correlation. We are experiments and how design strategies have Eric Beh now about five months away from the 9th been increasingly developed to capture the annual J. B. Douglas Awards, so I would ever increasing complexity that exists in South Australia Branch News like all supervisors of postgraduate students studying choice behaviour. An individual’s to seriously considering nominating them response, or choice, of a particular question The Expected Value of Information, to take part in the event. Information is often determined by factors that are Trial Design and Decision Making in about the date and location of the Awards not reflected in a survey. Therefore it Health Technology Assessment and dinner will be advertised as the time becomes important to understand, not draws nearer. just how an individual chooses an item The first meeting of the year for the On the 19th of March the Branch held or responds to a question but it is just as SA branch was held in February at the its AGM at the University of Sydney. important to understand why they make University of Adelaide. The speaker was Amongst many things on the agenda, a such choices. David demonstrated this Simon Eckermann who is currently an proposal was unanimously accepted by using several very interesting examples, Associate Professor in Health Economics those in attendance to modify the rules including what factors determine ones use at Flinders University and is also a which involved restructuring the makeup of public or private transport (cost, toll, member of the Economic sub-committee of the Branch council. Gone now are the traffic, etc), to better understand why such of the Pharmaceutical Benefits Advisory days where the NSW Branch will have a choice is made. To reveal the extent to Committee (PBAC). the outgoing President continue their stint which is occurs, David employed ordered Economic analysis is used in health on council for another two years as Past heterogenous logit models to provide a technology assessment to inform decision President. Instead, in the first year of a quantitative analysis of his data. makers of the expected cost of a new

SSAI Newsletter – June 2008 17 Branch Reports Continued

treatment relative to the cost of an existing (information button), users can obtain treatment. Ideally the new treatment will images of the census question, definitions be more effective and less costly than of variables, classification schemes and the existing treatment, however often it notes on the quality of the data associated is more effective but also more costly. with the variable. A decision must then be made about Prior to this Census, most tables of whether to implement the new treatment data prepared by the ABS were based on or not. By using patient level data, the “place of enumeration” (i.e. the location bivariate distribution of the cost and where the census form was completed). In response to user requests this has been effectiveness can be estimated through the changed so that most tables are now based use of bootstrapping. on “place of usual residence”. However Simon presented models for choosing Michael warned users to be cautious when between three alternatives when considering comparing results from the 2006 Census introducing a new treatment: delaying with earlier years, and to check that they are Michael told us how the ABS has acceptance of the new treatment while a making appropriate comparisons. He also put considerable effort into increasing pointed out the inclusion of mesh blocks randomised controlled trial is conducted, the accessibility of its data and statistical for the 2006 Census data which allows the accepting the new treatment and conducting information since the 2001 census, especially user to make almost perfect fitting of the a randomised controlled trial concurrently, in relation to the web. To use Michael’s data to the geographical areas they require or accepting the new treatment without words, they want to “let people find the such as suburb boundaries. conducting a randomised controlled trial. data before they die!”. Their goal has been The methodology was illustrated in the to remove the barriers between people and In the last part of the talk, Michael context of early and late interventions for the data. They have also recognised that explained that the consultation phase for pregnant women whose babies are in the people have different data needs - some the 2011 Census was starting in November 2007 and those interested in making a breech position, where the outcome of people require data that is customised for submission will have until March 2008 interest is whether the woman gives birth by particular geographic areas or on specific to do so. A number of the attendees caesarean or not. He then discussed a fourth topic, while others’ need are much simpler. Consequently, the ABS has produced a continued their discussions with Michael option when there are multiple jurisdictions at a nearby restaurant. involved, where one jurisdiction may use range of products to meet these needs. In conclusion, Michael’s talk was well a side payment to influence trial design To date, the ABS has four web based received, as highlighted by significant in another jurisdiction. In this scenario of products available (Quick Stats, Map Stats, Census Tables and Community Profiles), audience comments and questions multiple jurisdictions, it may be optimal for throughout the session. On behalf of the with two additional products due for some jurisdictions to adopt the treatment Statistical Society I would like to thank release early in 2008 (CData Online and with no randomised controlled trial, while a Michael for his lively and informative Table Builder). Michael provided us with trial is conducted in other jurisdictions prior interactive presentation and I encourage a demonstration of some of these products to their acceptance of the new treatment. everyone to visit the ABS website to find (for details I refer you to the ABS website out a few interesting facts about their Lisa Yelland www.abs.gov.au). neighbourhood, because it really is as easy Victoria Branch News as Michael says!! Helen Bartley October 2007 meeting Fisher and Scores Speaker: Michael Beahan The Belz Lecture, followed by the Belz Director of Census Output for the ABS Dinner, is the most significant event of the Topic: Dissemination of data from the In his demonstration Michael highlighted year on the Victorian Branch calendar. The 2006 Census the different ways that people can start lecture honours Professor Maurice Belz, looking for data – either geographically or by the Foundation Professor of Statistics at On Tuesday 16th October 2007, the topic. One of the exciting geographic search the University of Melbourne. The first Belz Victorian Branch of the Statistical Society features is that users can type in a specific Lecture was held in 1969, and in recent met at Swinburne University. Over 20 address (e.g. their home address), and the years it has assumed the additional status of people were present to hear our October search facility will find all locations (e.g. a University of Melbourne Public Lecture. guest speaker, the Director of Census census collection district, local government The Belz Lecture for 2007 was delivered Output for the ABS, Michael Beahan, who area, state) containing the property address on November 13 by Professor Alan Welsh. discussed the dissemination of data from for which census products are available. Alan is a well-known member of the the 2006 Census and provided us with Another new feature of Census 2006 is statistical profession. He is a Fellow of some brief insights into the planning for the metadata that is available to support the the Institute of Mathematical Statistics, the 2011 Census. tables. By the simple click of an “i” button a Fellow of the American Statistical

18 SSAI Newsletter – June 2008 Branch Reports Continued

Association and a Fellow of Australian response and the explanatory variables, consist of Mervyn Silvapulle, who will Academy of Sciences. that the correlation matrix between the continue a two-year term as President, response and the explanatory variables is and the following members, elected at not significantly different from having the AGM: Ian Gordon (Vice President), rank one, and that the estimated scores Owen Jones (Treasurer), Lucy Busija are significantly different from the equally (Secretary), Fiona Beer (Young Statisticians spaced scores (0, 1/4, 1/2, 3/4, 1). Bartlett Representative), Muhammad Akram, Kym computed marginal confidence intervals Butler, Chris Lloyd, Debra Partington and for the components of the score vector and Brian Phillips. The President extends his used these to argue that the assigned values thanks to retiring Branch Councillors Ann of 1/2 and 3/4 are inadequate. This last test Maharaj, Kris Jamsen, Kay Lipson and reaches the opposite conclusion to that of Geoff Laslett. Geoff, though no longer Fisher. Therefore, it would be of interest to on the Branch Council, will continue as compare the two analyses carefully and see Branch Editor of the SSAI Newsletter. At if we can find an explanation. the end of the meeting, Mervyn Silvapulle The tests of the first hypothesis (of no presented a eulogy for Chris Heyde. relationship or of nonzero correlation) turn Lucy Busija out to be based on equivalent statistics. Professor Alan Welsh Fisher’s F sampling distribution is arguably conservative compared to Bartlett’s chi- Statistical modelling of assessment and education data from a progress The title of Alan’s presentation was squared but these are both asymptotically “Fisher and Scores”. The talk described the same. It is therefore not surprising monitoring perspective work, motivated by G. K. Watson and that Fisher and Bartlett reached the same Following the Annual General Meeting, carried out jointly with John Robinson, to conclusion. H. O. Lancaster later showed Dr Siek Toon Khoo addressed the meeting, understand an analysis of R.A. Fisher and that the assumed chi-squared sampling attended by more than 30 people. The compare it to a contradictory analysis of the distributions of these tests are incorrect speaker is the Director of the Psychometrics same data by M.S. Bartlett. The data are a but Fisher and Bartlett were in agreement. Institute at the Australian Council for table of results obtained by G. L. Taylor by Fisher did not test Bartlett’s second Educational Research in Victoria. She testing 12 blood samples against 12 sera and hypothesis (the correlation matrix has rank presented an absorbing talk highlighting a recording the response as one of 5 ordinal one) but Bartlett’s test was not significant number of difficulties associated with using values. Fisher’s approach to analysing these so he also proceeded to test the adequacy classical ideas, especially Classical Test data was very similar to that used by him of the equally spaced scores. Detailed Theory, for modelling the measurement for linear discriminant analysis. He used asymptotic expansions show that both of achievement growth in education. test statistics have asymptotic chi-squared the analysis of variance to show that there Her talk was based on the results of distributions (although Fisher’s test uses far is a significant relationship between the the Australian Council for Educational more degrees of freedom than Bartlett’s) but response and the explanatory variables (sera Research Longitudinal Literacy and that they are testing different hypotheses. and blood samples) and the analysis of Numeracy Study (LLANS). LLANS Essentially, Fisher’s test combines a test covariance to make approximate inference was initiated in 1999 with the aim of for the adequacy of the assigned scores about the scores (the vector estimated from studying achievement growth in literacy the data which assigns numerical values with another for the hypothesis that the and numeracy for Australian children, with to each ordinal category). He showed that correlation matrix has rank one whereas a particular emphasis on the longitudinal the estimated scores are not significantly Bartlett treated these hypotheses separately, approach in assessing learning skills in the different from the equally spaced scores using separate test statistics. early years of school. (0, 1/4, 1/2, 3/4, 1) and gave an implicit Mervyn J. Silvapulle expression for the acceptance region of the test which is a joint confidence region Victorian Branch Annual for the scores. It is interesting that from a General Meeting modern point of view, we can see that Fisher The Annual General Meeting was making inference about eigenvalues (AGM) of the SSAI Victorian Branch and eigenvectors. was held on Tuesday 18 March 2008 at Bartlett was critical of Fisher’s approach, the Hawthorn campus of Swinburne arguing that it used univariate methods to University of Technology. The Branch try to summarise a multivariate relationship. President presented the Annual Report He also felt that the analysis of variance and the Treasurer’s Report. In 2007, the tests were unjustified and that likelihood Branch held seven meetings at different ratio tests (justified by asymptotic results) locations, including La Trobe University, should be used instead. Bartlett carried Monash University, Swinburne University out three tests and showed that there of Technology and The University of is a significant correlation between the Melbourne. The 2008 Branch Council will Dr Siek Toon Khoo

SSAI Newsletter – June 2008 19 Branch Reports Continued

Siek Toon outlined the process of the development of meaningful assessment of Tea party held for t-test latent variables, which are unmeasurable through physical measurements. She explained that in order to show growth 28 March 2008: It was an unusual event. “Having the t-test means that I can look during the first three years of school, Singing happy birthday to an inanimate for significant differences in the data that measurement scales for literacy and object or dead person is generally deemed, researchers supply and I’m confident in numeracy needed to be constructed first. at the very least, unnecessary, or perhaps the conclusions I draw. If it wasn’t for the These scales were used to track growth in even uncouth, but on this particular t-test, we’d all be working ourselves to the literacy and numeracy achievement over occasion it was requisite for the party’s bone collecting large amounts of data, or time. The scales provided estimates of enthusiastic followers. else relying on less powerful tests.” students’ ability, allowing students to be The Faculty of Information Sciences Zoe Spinocchia positioned along the ability continuum at and Engineering played host to a party the time of assessment. celebrating the 100th birthday of the She discussed, using Rasch modelling t-test, the most widely used statistical of assessment data in constructing procedure for comparing two groups of measurement scales, how a number of observations. models were examined to assess the ability Head of Mathematics and Statistics of the students. These models identified Peter Vassiliou provided party attendees where students were underperforming with a brief background on the test and and needed additional work to help them the life of it’s founder William Gossett, improve. From this Siek Toon went on to underlining the test’s continuing relevance describe how the LLANS had collected a in all fields of research today. series of assessment data over time for each The test was developed in 1908 by individual student in the study. This led to William Sealy Gossett, who worked as the question: ‘How can you investigate the a statistician at a Guinness brewery in individual differences in developmental Ireland. He used the pen name “student” trajectories as the students grow in literacy to ensure that Guiness’ industrial and numeracy skills?’ The basic idea is to processes were kept covert. Today, the use longitudinal growth curve modelling “Student’s t-distribution” is routinely to model the developmental data. used to perform tests of hypothesis in all Siek Toon enthusiastically explained types of research. why the curves are best fitted by a two- Statistics Lecturer Alice Richardson stage model. During the first stage of was enthusiastic about the event. “The Information Sciences & Engineering data collection, the assessments were 100th anniversary of the t-test is definitely Research Associate Fariba Shadabi uses administered by teachers using one-on- something worth celebrating,” she said. the t-test to analyse data for her research. one interviews. The interview assessments were administered twice during the first and second years of school and at the beginning of the third year of school. The second stage of data collection started in the fourth year of school. The students completed a written assessment every year from the fourth year to the seventh year of school. The longitudinal data obtained across time allowed tracking of student performance growth. The growth modelling investigated individual differences in the developmental trajectories in the two stages of growth. Debra Partington

Statistical consultant Dr David Pederson, senior lecturer in Maths and Statistics Dr Shuangzhe Liu, Professor Alan Welsh of ANU, and lecturer in Statistics Dr Alice Richardson celebrate at the t-test party.

20 SSAI Newsletter – June 2008