ewsletter June 2007 Number 119 N statistical society of australia incorporated Young Statisticians’ Conference 2007
Although I have attended several conferences, this was the first one where there were so many people from my own age group – I wonder if the title “Young Statisticians’ Conference” had anything to do with it. I learnt a great deal from some of the senior statisticians who spoke. I particularly found Michael Adena’s talk interesting, as I never knew about GStat and AStat accreditation before. Terry Neeman’s talk on “statistics and ethics” made me think about what I would do if I had to choose between compromising scientific principles and keeping a client happy. It is easy for me to think I have high moral standards, because I am single, but would I be able to keep this standard if I had a family that relies on my income? Of course, I also enjoyed many of the talks given by young statisticians. I learnt that I should prepare for and expect the Young Statisticians on their Sunday tour following the Young Statisticians’ unexpected from Lisa Yelland, such as a Conference 2007. blown up letterbox that contained filled- in surveys, or that one of your study On the 13th and 14th of April, the As well as the keynote speakers, participants may end up in jail. One of my Young Statisticians section of the statistical participants were encouraged to present favourite talks was Damien Kennedy’s talk society held a conference at the Australian a ten minute presentation relating to on “Data Visualisation”, as this has always National University, Canberra. The theme the work that they do. Official statistics been an interest of mine. I was fascinated of the conference was “We’re young and (both from the ABS, and Statistics New by the resources on the web and the we count”. Exactly how important young Zealand), Academia and Industry were quality of the dynamic graphics. statisticians are to the discipline was all well represented, and the range of Of course, being a student I cannot shown by the array of presentations given topics was huge. The prize for the best forget to mention the conference dinner. by participants. talk was presented to Jason Rabbitt from Both the quality and quantity of food were Keynote addresses were made by some Data Analysis Australia who discussed excellent. Having finished my presentation prominent statisticians, each showing the his analysis of the water quality in the before the dinner, I was able to relax and wide variety of interesting paths that Cockburn Sound, where a desalination enjoy the food, with a few sips of wine are available to early career statisticians, plant has been commissioned. every now and then. and the genuine need for these skills in On the Friday evening, a conference Finally, I’d like to thank Jane Waslin official statistics, research, and industry. dinner was held in University House, and the conference organising committee These keynote speakers were the where we were all able to talk over a great for organising one of the best conferences Australian Statistician Mr Brian Pink, dinner, and even have a go at making I have ever attended. I would also like to Dr Amy Salter from the University of things from balloons. As well as the thank the Canberra Branch for sponsoring Adelaide, Mr Michael Johnson from conference dinner, the following evening me with a Student Grant. Roche Products, and Mr Neville Bartlett Kevin Wang from NR Bartlett consulting. continued on page 2 Young Statisticians Conference 2007
From page 1 Having the conference in Canberra was an advantage as we had easy access to many we hit the streets of Civic for dinner and government employers of statisticians drinks at King O’Malleys. Both of these who sent staff to the conference and evenings proved to be a great opportunity also came along and spoke at the careers to catch up with people we already knew, session at no cost to us. At the careers and also to get to know other people we session employers gave talks about their had just met a bit better. organisation and opportunities that exist. I would like to thank the organising The feedback from this session was really committee who made sure that the good, especially from students, who conference was a huge success: Janice really want to know about career options Wooton, Richard Hutchinson, Simon available. I would highly recommend Bartlett, Theresa Nunan, Penny Bennett, having a careers session at the 2008 Lisa Yelland, Pam McCaskie, Kevin Australian Statistical Conference. On Wang, Kris Jamsen, Liz Jermyn and Jane the downside of having the conference Waslin. Thanks also to the sponsors: ABS, Rhianon Marchant receiving her GStat in Canberra there was some criticism CSIRO, Roche Products, Covance, and Certificate at the conference. that the conference did not have a large Data Analysis Australia. commercial presence. This is something conference of this size nor the time In short, I thought that the 2007 Young we will keep in mind for next time. to do so (all are studying or working Statisticians conference was a great success, full time). Without Jane and Liz’s help The conference program over the and I hope that we can meet again in the conference would not have been as two days was very full. We managed to two years time for an equally successful professional looking as it was (unless fit in all the talks that were submitted conference. In the meantime, I hope that we had hired a professional conference before the deadline, but this meant that all early career statisticians get involved in organiser at considerable cost). We thank contributed talks could only be about the young statisticians section, because we Jane and Liz for all of their effort. 10 minutes. This was commented on as are young, and we do count. Having Jane and Liz to handle the being quite short and rushed and it was Stephen Bush administrative aspects of the conference suggested that next time we should have As a member of the organising committee allowed the organising committee to parallel sessions. It is nice to have this for the Young Statisticians’ Conference put together a program and make other problem to think about as a result of so 2007, I would like to share some reasons for decisions in line with their expertise. The many people wanting to present. the success of the conference. organising committee had representation Summing up, I think the Society Jane Waslin (Executive Officer of the from New South Wales, Victoria, should aim to have a Young Statisticians SSAI) and her assistant Liz Jermyn took Western Australia, South Australia and Conference every 2 years, during the ASC care of all the administrative work in the ACT. Each representative promoted off-year. There was a lot of enthusiasm organising the conference from the SSAI’s the conference in their state/territory and energy from young statisticians during office in Canberra. Having Jane and Liz through their local Branch as well as by this year’s conference and it would be nice to contact and liaise with sponsors and putting the word out to key organisations, to keep this momentum going. keynote speakers, produce the poster and such as universities. At final count, there On the Sunday after the conference registration form, organise the venue, were 88 delegates from every single a small busload of delegates toured handle registrations, create and update state and territory of Australia as well as Canberra for a few hours of sightseeing the website, and all the other tasks I am New Zealand. This great turnout would and networking. sure I have missed, was invaluable as not have been achieved without having the rest of the organising committee had committee members from right around Richard Hutchinson no experience in organising a national Australia acting as local contacts. [email protected]
In this issue Young Statisticians Conference 2007 1 Environmental Statistics Section 8 Member News 3 Annual General Meetings Notice 9 Editorial 4 Obituary 10 President’s Corner 5 Biometrics by the Beach 11 Statistics Undergraduate Vacation Program 6 CensusAtSchool International Workshop 12 Three Doors 8 Branch Reports 13
2 SSAI Newsletter – June 2007 Member News
Eugene Seneta receives the 1996) an ingenious method for visualising Today?” In his 30 minute address, Peter Hannan Medal in 2007 conditional probability densities, and gave persuasive evidence that today’s Eugene Seneta has done much seminal (in the American Statistician in 1996) a statisticians live in interesting times. work in probability and statistics in remarkably useful tool for data analysis— New technologies have driven innovation connection with Markov chains and the “highest density regions’’ boxplot. The in statistical methodologies. Computing non-negative matrices, branching latter is used today in forecasting, when power allow us solve much more the forecast densities are multimodal (as processes, the history of probability and complex problems than were possible 50 can occur in nonlinear forecasting), and statistics, and in such diverse areas as years ago, but also new technologies are in Markov Chain Monte Carlo studies. slowly varying functions, Bonferroni type creating new types of data which demand bounds on probabilities of unions of sets, Peter Hall new statistical techniques. Units of data on modelling of the price of a risky asset, from olden times were predominantly and in the scaling of Higher School New Fellows Day at the Australian single values, or small vectors of values. Certificate marks. The implications of Academy of Science Today’s data units include may be large some of his research are considerable. The Science at the Shine Dome is an annual dimensional vectors, functions and variance-gamma finance model, proposed three-day event in early May celebrating surfaces. Peter gave two examples of by Madan and Seneta in 1990 is in use on the achievements and advancements in how technology has driven methodology. Wall Street. The algorithm which Seneta science by Australian scientists. These His first example was in the analysis produced for scaling Higher School three days bring together newly elected of microarray data, which requires the Certificate marks in the early 1980 was Fellows of the Australian Academy of consideration of thousands of correlated later used to determine the New South Science, early career researchers, science hypothesis tests. His second example Wales Tertiary Entrance Rank. Seneta teachers, and current Fellows of the was in the development of sensitive is a scholar of the highest international Australian Academy of Science for methods for detecting weak signals, such calibre and is very well known for his scientific seminars, professional training as testing for evidence of bioweapons in contributions. and award presentations. The highlight the atmosphere. of the event is New Fellows Seminar Day The Hannan Medal of the Australian when the 16 newly elected Fellows of the Academy of Science recognises the Australian Academy of Science, each of achievements in time series analysis of the them distinguished in their respective late Professor E.J. Hannan, FAA. Because fields, give short presentations on their of Professor Hannan’s broad interests life-time work. This year’s series of ten in the mathematical sciences the award minute talks were simultaneously short is made in one of three areas in turn at and profound, representing seminal two-yearly intervals – Pure Mathematics; contributions of the new Fellows in the Applied and Computational Mathematics fields of medicine, molecular biology, plant and Statistical Science. biology, climatology, chemistry, statistics, Chris Heyde optics, nanotechnology and information science. It was no small coincidence that Rob Hyndman receives the the first speaker and the final speaker of Moran Medal in 2007 this fascinating lecture series were two of Australia’s distinguished statisticians. Rob has made major contributions Alan Welsh and Luke Prendergast to a wide range of fields, especially to forecasting, time-series, graphical methods, and methodology based on Professor Alan Welsh, newly elected conditional density estimation. In Fellow of the Australian Academy of forecasting his research includes that Science, closed the New Fellows Seminar in his 1997 article in the Journal of Day with some incisive comments on the Royal Statistical Society, challenging the modelling and analysis of clustered the appropriateness of the most data. Clustered data arise in many fundamental of Bayesian forecasting contexts in both the experimental models for exponential-family time and the survey setting. Alan reviewed series; and his seminal paper on state- different possible models for clustered space models for exponential smoothing, data, explored differences between them published in the International Journal of and discussed the implications of these Peter Hall (centre) Forecasting in 2002. Both articles have differences. He closed with an example very important ramifications for both of the ecological fallacy of interpreting statistical methodology and practical Professor Peter Hall, recipient of the associations observed at the aggregate forecasting. Rob’s research on graphical 2007 Matthew Flinders Medal, opened level as associations at the individual methods has produced (in the Journal of the ceremony with his Matthew Flinders level. Computational and Graphical Statistics in Lecture titled “What Excites Statisticians Terry Neeman
SSAI Newsletter – June 2007 3 Editorial
Once again this issue of the newsletter < 0.999 in steps of 0.001. David (2005) PO Box 5111, is a chance to celebrate the achievements reports that “Sheppard did a splendidly of the people in the Statistical Society. accurate job, apparently no errors ever Braddon ACT 2612 Special congratulations go to Rob having been reported.” His figures form Phone (02) 6249 8266 Hyndman, Eugene Seneta, Peter Hall and the basis of countless versions of tables of Fax (02) 6249 6558 Alan Welsh, who have received awards the Normal distribution, generally without Email: [email protected] and honours from the Australian Academy any reference to the original. The work is Society Web Page of Science. You’ll also find reports from clearly so fundamental that authors feel http://www.statsoc.org.au the organisers and attendees of the Young that it must have always existed! Statisticians Conference, who travelled Sheppard himself died in 1936 from far and wide to Canberra in April; the final report from the outgoing President, and obituaries were printed in several Editors Kaye Basford; and all the usual Branch places, including the Annals of Eugenics. reports, for news from close to home. R.A. Fisher, editor of the journal, also Alice Richardson, School of ISE, contributed a note on Sheppard’s work. University of Canberra, Also, this month let’s celebrate the One hundred years of continuous use PO Box 1, Belconnen ACT 2616 centenary of one of the fundamental Φ-1 ζ tools of statistics, namely the table of the of ( ) is a fine achievement indeed. Normal distribution, and the Australian Although many statisticians would now Michael Adena, Covance Pty Ltd connection that goes with it. In June reach for an electronic solution to normal PO Box 5125, Braddon, ACT 2612 1907, Biometrika printed an article entitled distribution calculations, the portability “Grades and Deviates: (Including a Table and “big picture” provided by a printed Correspondence of Normal Deviates Corresponding to table of values surely still has a place in Please direct all editorial Each Millesimal Grade in the Length of statistics teaching and practice. correspondence to Alice Richardson. an Array, and a Figure).” The author of References Email: [email protected] the article is given as Francis Galton, but the section at the end entitled “Table of David, H.A. (2005). Tables related to the normal distribution: a short history. Disclaimer Deviates of the Normal Curve” is credited The American Statistician 59, 309 – 311. The views of contributors to this to W.F. Sheppard. Galton, F. (1907). Grades and Deviates: Newsletter should not be attributed to William Fleetwood Sheppard was the Statistical Society of Australia, Inc. born near Sydney in 1863 and went to (Including a Table of Normal Deviates Brisbane Grammar School followed by Corresponding to Each Millesimal Grade in the Length of an Array, and a Figure). Subscriptions Trinity College, Cambridge. He met Francis Galton in Cambridge, and Galton Biometrika 5, 400 – 406. The Newsletter of the Statistical encouraged Sheppard in his statistical Society of Australia is supplied free to Wikipedia (2007). William Fleetwood endeavours. all members of the society. Any others Sheppard. [online] accessed 26/3/07. wishing to subscribe to the Newsletter The 1907 paper gives values, to four http://en.wikipedia.org/wiki/William_ Φ-1 may do so at an annual cost of decimal places, of | (z)| for 0.000 < z Fleetwood_Sheppard A$30.00 for an issue of four numbers.
Advertising Advertising will be carried in the Conferences Newsletter on any matters which ISI 56th Session the Editors feel are of interest to 22-29 August 2007, Lisbon, Portugal the members of the Society. For http://www.ine.pt details of advertising rates, etc. contact the SSAI Executive Officer at ICAS-4 – Advancing Statistical Integration and Analysis, Fourth [email protected] International Conference on Agricultural Statistics 22-24 October 2007, Beijing, China Printer http://www.stats.gov.cn/english/icas National Capital Printing IBS Australasian Region, ‘Biometrics on the beach’ 22 Pirie Street, Fyshwick ACT 2609 2-6 December 2007, Coffs Harbour, NSW http://www.biometrics.org.au BioInfoSummer07 DEADLINE FOR ICE-EM Summer Symposium in BioInformatics NEXT ISSUE: 10-14 December 20076, Australian National University, Canberra, ACT. 10 August 2007 http://www.maths.anu.edu.au/events/BioInfoSummer07/
4 SSAI Newsletter – June 2007 President’s Corner
In my previous newsletter contribution, I commented on the improvement in funding, National Strategic Review of Mathematical Sciences Research in not just to the elevation to Australia which followed the Review on Statistics at Australian a higher band.) Universities. In the latter, the recommendations were broadly This is an excellent grouped around issues relating to the school programme, result which will have a organization and funding within and among universities, and major impact on those image and profile of statistics (including employer/university statisticians working in interactions). The recent review found that the mathematical tertiary institutions, with sciences in Australia require an immediate and substantial resulting positive benefits capital injection to build a critical mass of research, education, for their students. I would industry and government interaction, and ensure we maintain particularly like to thank our technical and problem-solving capability. all those who contributed to the reviews and The outcomes of the two reviews were consistent and assisted in the extensive reinforced the way forward for statistics in this country. Hence, I promulgation of their am very pleased to report that these reviews have contributed to recommendations. a major change in the way mathematics and statistics are viewed Finally, I would stress (along with Peter) that we continue by those in government. In doing so, I shall quote from an e-mail to strengthen the cooperative spirit that we have developed message written by Professor Peter Hall, an eminent statistician among those working across all levels of education (be it schools and current President of the Australian Mathematical Society. or universities), research entities, government institutions and As many of you will have heard, the federal government’s budget private enterprise. announcements last Tuesday promise to significantly improve the This will be my last President’s Corner. By the time the next state of the mathematical sciences in Australia. The universities and newsletter comes out, William Dunsmuir will have taken over. CSIRO will benefit broadly from budgetary measures announced I have confidence that he will do an excellent job in leading by the government, but more specifically, the fields of mathematics our professional association and wish him all the best in that and statistics in Australian universities were singled out for special, endeavour. It has certainly been an exciting and productive two positive treatment. Those discipline areas will be placed into a higher years for me. Thanks for that honour and privilege! funding band, shared with computer science and resulting in an increase of a little over 50% in financial support per mathematics Kaye Basford and statistics student. (Some of this increase is due to a general E-mail: [email protected]
Australian Statistical Conference 2008 30th June to 3rd July 2008, Sofitel Hotel, Melbourne
Main Themes and Organisers • Environmental Statistics Keynote and invited speakers will be listed Organisers: Bronwyn Harch, Geoff Laslett, Petra Kuhnert in later announcements and on the ASC2008 web-site. • Medical Statistics Organiser: John Carlin Conference web-site: Go via – http://www.statsoc.org.au/asc2008 • Stochastic Modelling and Statistics in the Social Sciences Will include E.K Foreman Lecture and sessions on longitudinal surveys and data, survey sampling, census data enhancement, financial statistics and more. Organisers: Ross Maller, Alan Herning, Geoff Lee, Veronica Program committee: Michael Martin Rodriguez Ross Maller (chair) Steve Roberts • Statistical Genetics and Bioinformatics Geoff Laslett Local organising committee: Organiser: Sue Wilson Bronwyn Harch Neville Bartlett (chair) John Carlin • Young Statisticians Kay Lipson John Maindonald Sessions will include one entitled ‘experiences of early Brian Phillips Geoff Lee career statisticians’ and another on ‘potential career paths for Katrina Scurrah Petra Kuhnert statisticians’. Derchieh Hung Organiser: Theresa Nunan Alan Herning Veronica Rodriguez Jane Waslin • Statistical Methodology Sue Wilson Conference organiser: Organisers: John Maindonald, Michael Martin, Steve Roberts Theresa Nunan FoxEvents, Melbourne
SSAI Newsletter – June 2007 5 The Statistics Undergraduate Vacation Program 2006/07: A Pilot Project
A new initiative tried in South statistics rather than food, the project an interview just after the examination Australia is proving to be a great success. has still given him an edge over other period finished, for a potential start The statistics undergraduate vacation graduates without such practical the following week. The participating program aims to give students real-life experience. organisations were flexible with respect practical experience in the application This experience has made me think to vacations that had already been and reporting of statistics. As the about my own career development and booked by students. Two applicants were initiator of this program I would like to those of my colleagues. I was fortunate to interviewed and both were offered one of share my experience. work on an industry-sponsored honours the projects. Unfortunately, one student project, which included work experience. received a better vacation work offer Background I have since come to the conclusion the and pulled out after initially accepting I work as the senior statistician within vacation work in the mathematical and the offer. the Food Safety Research Program of statistical sciences is more of a novelty the South Australian Research and than the norm. In contrast, engineers Outcome and what we’ve learnt and architects, for example, would be Development Institute (SARDI). In my Lu Li, a second year finance / statistics unlikely to gain employment without position I work with R&D companies, student, worked on a project to identify vacation work experience. In fact, some such as Meat & Livestock Australia factors related to rheumatoid arthritis (MLA), government agencies and food engineering courses included vacation progression, jointly funded by the companies on all kinds of food and food work as a pre-requisite for graduation. Early Arthritis Clinic (Royal Adelaide safety related projects. Getting the program off the ground Hospital) providing the medical During the university summer break supervision, and SARDI Food Safety of 2005/06 I was involved as a mentor/ Consequently, when I was elected vice- providing the statistical supervision. president of the SA branch of the society supervisor in MLA’s undergraduate During the vacation work, Lu learned in early 2006, I decided that it would be vacation program which aims to: many important skills – statistical, worth a try for the branch to take a lead • Make professional undergraduates project management, communication role in organising a vacation program of aware of opportunities within the and presentation skills. In addition, some kind. After all, we’ve got members Australian red meat industry she learned how important and time- in many industries, government areas consuming data cleaning can be; • Give companies access to the and universities, and most of them, I knowledge that professional people expect, would have some data sets which finalising a clean data set took about 50% from many backgrounds can haven’t been fully explored. However, of the project’s allocated time. At the provide in order to pilot the idea I didn’t want end of the project, Lu prepared a short • Give companies an opportunity to to go too big to start off with. So, after project report and presented her results “interview” potential candidates in initial discussions by the branch council, to researchers from the Early Arthritis a real life situation over a three- we decided on an approach which would Clinic. In addition, the SA branch month period. help develop projects and facilitate invited Lu to present her experience Clearly, MLA has much greater the linking of students with projects, with the vacation program at the 2007 financial resources than SSAI, which without taking a role of responsibility AGM – her comments confirmed to me is reflected in the founding of vacation for the selection of applicants. that this was a worthwhile endeavour. projects. Students are paid $6,000 for Once the design was sorted out, I While I think that the 2006/07 12 weeks of work, travel to and from the got in touch with some of my personal undergraduate vacation program was a participating company (including a trip contacts to investigate the opportunities success, there are of course things that home at Christmas) and accommodation for projects. Several indications of I think we can do better. These include for two weeks. In addition, mentors interest materialized into two projects. the following: are paid for their time and travel The biggest hurdle was the financial • Projects need to be identified and requirements to meet with students remuneration of students. My initial scoped much earlier. This will allow on a fortnightly basis. Consequently, ideas, which were based on my MLA for earlier advertising of projects as total project budgets are typically about experience, were in the end substantially well as better data cleaning prior to $20,000-$25,000 – funded half by MLA off the mark – the final amount was a student starting. and half by the participating company. $1500 (for 8 weeks). • Projects and associated data need to I found the experience very rewarding We advertised the two projects, as be ready when the student is ready and I believe that the company I was well as the application process, directly involved with made significant head-way to students by mid-October through to start. This will result in less time in their understanding of the slaughter lecturers at The University of Adelaide being wasted by continually having process. In addition, the student gained and the University of South Australia. to clean data. valuable industry experience. While the The closing date for applications was • Projects need to be advertised to student has decided that he would like set to 3 November (at the end of swot- students much earlier – before to pursue a research career in sports vac) and students were called in for they make alternate vacation plans.
6 SSAI Newsletter – June 2007 This will hopefully result in more • A society which provides a real and applications. tangible benefit to student members, Society Secretaries • Students may need considerable which should ultimately help with member retention. Central Council encouragement in order to apply for President: Prof Kaye Basford a project – a question-and-answer Secretary: Dr Doug Shaw session with students early in the Where to from here? [email protected] application process may be useful This exercise in leadership, undertaken Canberra for this. This will hopefully also by the SA branch, has highlighted the President: Ms Glenys Bishop result in more applications. effort required to bring together and Secretary: Dr Ray Lindsay coordinate all the players – students, [email protected] • Students should be required to lecturers, society members, organisations, New South Wales submit weekly project reports to the potential funding bodies and possibly President: Ms Caro Badcock branch representative. These reports others. Some of the participants may Secretary: Dr Eric Beh should include what the student not be experienced in dealing with each [email protected] has done during the last week, other, which can lead to a potential Queensland what problems were encountered, President: Ross Darnell conflict in expectations. Since it is the Secretary: Ms Helen Johnson and how the student is planning society who initiates these collaborations, [email protected] to address those problems. These it is the society’s role to be on top of
reports will result in a record in the potential conflicts so that they can be South Australian case of a conflict. President: Dr Andreas Keirmeier mitigated before getting out of control. Secretary: Dr Paul Sutcliffe • Aim for projects to pay students It is also important that we, as a [email protected] more than $1500 for 8 weeks. council and as a society, recognise Victoria Greater remuneration will make that the program aims to facilitate the President: Professor Mervyn Silvapulle statistical projects more attractive. process, but that we do not want to take Secretary: Dr Ann Maharaj The advertising process has also direct responsibility for outcomes that [email protected] highlighted that considerable support should and are not controlled by us – all Western Australian from lecturers is needed. They have direct the fun without the responsibility. For President: Dr Brenton Clarke contact with students and consequently example, I think it is reasonable for the Secretary: Ms Pamela McCaskie [email protected] are in an ideal position to push the value society to help advertise the program, of such practical experience. Without and maybe a councillor should even be their support, students are unlikely to part of the selection panel. However, Section Chairs the selection of a suitable candidate be interested. Bayesian Statistics should ultimately rest with the funding Also, from an organisation point Kerrie Mengersen organisation – they are putting up the of view, I now believe that a certain [email protected] money and they need to be comfortable separation needs to be maintained Biological Sciences with their choice. This, I hope, also between the program and projects. For Ari Verbyla reduces the amount of work required by [email protected] the next iteration I am planning to council members. be involved in overseeing the program, Environmental Statistics The SA branch will be running the Petra Kuhnert – [email protected] but I prefer not to be involved as a vacation program again this year – with project supervisor – roles get muddled Industrial Statistics some changes. We are trying to identify too easily. Ross McVinish – [email protected] three to four projects for this year’s Clearly, the idea of vacation work is by Statistical Computing vacation program much earlier – we Kuldeep Kumar no means unique. Some of our universities are currently in the process of doing [email protected] have already invested resources in making this. We are also planning to inform Statistical Education (Co-Chairs) similar programs happen. However, I students about available projects much Michael Martin think that our society, with its diverse earlier – beginning of second semester [email protected] membership, is in a unique and ideal – with the hope that more students will Peter Howley position to take a leading role to develop apply for the projects. In addition, I am [email protected] and drive structured vacation programs aiming to visit some statistics classes to Surveys and Management aimed solely at statistics graduates. The advertise the program personally, answer Veronica Rodriguez potential benefits include: students’ questions and encourage more [email protected] • Statistical graduate students with students to apply. Young Statisticians Janice Wooton real experience I would be happy to discuss SA’s [email protected] • A better profile for the profession in experience with the program with any of the society’s members. I am Social Sciences general Michele Haynes – [email protected] particularly interested if you have • A linking between organisations ideas of how to improve the program Further contact details for and existing members which can further or have a project for a student Society Secretaries and lead to additional collaborative to work on during the 2007/08 summer Section Chairs can be obtained relationships period. You can contact me either via by contacting the Society on • A society which is seen as progressive [email protected] or (02) 6249 8266 and responsive to members needs on (08) 8207 7884.
SSAI Newsletter – June 2007 7 Three Doors with Borek Puza (Edition 10)
Welcome to the 10th edition of Three Doors. Last time I presented However, it may be counter-objected that, with d=1 (say), the The Two Envelopes Puzzle and am now pleased to announce Terry above logic allows for the other envelope to contain 50, 199, 200 Neeman as the latest winner of The Three Doors Prize, a cheque or 201 dollars, in contradiction of the premise that one envelope for $60 as donated by the SSAI. We now present the solution, and contains twice as much money as the other. If we understand “twice” the next puzzle follows. to mean “exactly twice”, then it seems that R was first rounded to yield the smaller amount M, and then the larger amount was taken The Two Envelopes Puzzle as exactly 2M. But this logic leads us back to the discrete case and the solution p=1/2. Or does it? Consider two envelopes such that one contains twice as much money In fact, there are an infinite number of possible solutions, since the as the other. The smaller amount in dollars was obtained by randomly nature of the rounding was not specified in the puzzle. For example, drawing a number from the uniform distribution between 0 and 1000. suppose that R~U(0,1000) was rounded to the nearest unit if less You now randomly select an envelope and find $100 in it. What is the than 100 and otherwise to the nearest multiple of 10, and then the probability that you have selected the smaller amount? result was doubled to yield the larger amount. Then X=100 implies either A={M=100}={99.5 8 SSAI Newsletter – June 2007 NOTICE The Annual General Meetings of The Statistical Society Of Australia Inc and The Australian Statistical Publishing Association Inc. will be held on Tuesday 31 July 2007 at 4.30 pm at the Australian National University, Canberra – room location will be posted to: http://www.statsoc.org.au/WhatsNew/ by 30 June 2007. SSAI Annual General ASPAI Annual General Meeting – Agenda Meeting – Agenda 1. Apologies and Proxies 1. Apologies and Proxies Proxies must be given in writing as per form inserted in Proxies must be given in writing as per form inserted with June the June 2007 issue of SSAI Newsletter. Proxy forms must 2007 issue of SSAI Newsletter. Proxy forms must be received be received by the SSAI Executive Officer for passing to by the SSAI Executive Officer for passing to the Secretary no the Secretary no later than 24 hours before the time of the later than 24 hours before the time of the meeting. meeting. 2. Confirmation of the minutes – Minutes of the 2. Confirmation of the Minutes – Minutes of the meetings as circulated meetings as circulated 3. Matters arising 3. Matters arising 4. Presentation of the Annual Report by the Editor 4. Reports of the Australian and New Zealand Journal of 4.1 President Statistics 4.2 Treasurer 5. Presentation of the Annual Report by the 4.3 Branches Newsletter Editors 4.4 Sections 6. Treasurer’s Report 5. Accreditation 7. Appointment of signatories 5.1 Report from Accreditation Committee 8. Special Business 6. Conferences 6.1 ASC 2008 (joint with NZSA) 9. Other business 7. Election of Section Chairs 10. Time and place of next meeting. Nominations for Section Chairs should be received at the SSAI office no later than 24 July 2007. Nomination Forms have been inserted in each copy of the June 2007 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. APBG could be considered a success. The workshop Bayesian approaches and the idea of double Australian Pharmaceutical was attended by over 70 people from the comparators. The lively discussions indicated the need for more development of these Biostatistics Group pharmaceutical, government, health care and academic sectors. It included statisticians, approaches and greater understanding to Indirect Comparisons in Health health economists, health outcomes scientists extensions of the topic, for example Mixed Assessments Workshop and university lecturers. The topics ranged Treatment comparisons. Full details of the If the relationship between attendance from a basic introduction of the concept of presentations can be found on the APBG and success is highly correlated, then indirect comparisons and its applications to website: www.apbg.org.au March’s workshop on Indirect Comparisons difficult issues like random effects models, Alan Brnabic SSAI Newsletter – June 2007 9 Obituary Rupert Thomas Leslie 1916–2007 Rupert Thomas Leslie died in Glasgow, Scotland on Good Friday, April 6th 2007. He was ninety years of age, and had been in poor health for several months. Rupert Leslie was born in Victoria on August 12th 1916, and received all of his education in Melbourne. He completed a BSc in Physics in 1938, and a BA in Mathematics in 1939, both at the University of Melbourne. After a couple of years of teaching Physics in Melbourne (Melbourne Grammar) and Sydney (Cranbrook), he became in late 1941 an Assistant Research Officer in CSIR, initially in the Physics Section of the National Standards Laboratory in Sydney. His starting salary was £366 per annum. He held this position until 1944, then transferred to the Section of Mathematical Statistics. In 1945 and 1946 he was with the Mathematical Statistics Section located Rupert Leslie (centre) in 1968, with Doug Shaw (left) and David Culpin. at the Division of Plant Industry and Economic Entomology in Canberra, In 1960, Rupert re-joined the, by now Rupert continued to teach at Strathclyde where he worked with George McIntyre. re-named, Division of Mathematical until 1981; on his retirement he was In 1946 he returned to Melbourne, Statistics in Sydney, becoming the Officer awarded an Emeritus Professorship. working as a Research Officer in the in Charge of the Sydney group. He However, teaching was in his blood, and Mathematical Statistics Section at the consulted extensively with researchers in he taught Applied Mathematics to final Division of Forest Products with, amongst the Divisions making up the National year students at a Glasgow Grammar others, Evan Williams. Measurement Laboratory, and he was School until 1994. In 1950, he joined the faculty of highly regarded within the Laboratory for Melbourne University, and, while there, Rupert was a strong supporter of the his skill in producing practical solutions Statistical Society in the 1960s. He was was awarded an MA in 1951 for a thesis to statistical problems. He was also well on “Mercerian Theorems”, and a PhD in the first President of the NSW Branch in aware of the human interaction aspects 1963 and 1964, after the formation of the 1958 for a thesis on “Statistical Treatment of consulting. Young statisticians in his Australian Society. He also put a great of Observer Effects in Scientific group who were about to go and meet deal of effort and organisational ability Experiments”. a researcher to discuss the researcher’s into NSW Branch symposia. Lunchtime One of Rupert’s stories was of being statistical problems would get a briefing visitors to the Division’s office in Alpha left alone with Fisher when the great from Rupert. This wouldn’t be about how House, Newtown would sometimes find man visited Melbourne University, and to solve the statistical problem; it would be Rupert and other staff members walking having to start a conversation. As Rupert about the researcher’s interests (scientific round and round the table tennis table, told it, the interaction went something and otherwise), his idiosyncrasies, his collating roneoed pages of symposium like this: relationships with others in the laboratory, papers. Leslie (brightly): I believe you knew and so on – necessary background in Rupert Leslie was an excellent applied Hardy, Professor Fisher? establishing a relationship! statistician, and an active proponent of Rupert spent part of 1968 and 1969 Fisher (huffily): Yes. Wrote a book the discipline and the profession. He called “A Mathematician’s Apology”. on leave at the University of Kentucky, then in 1970 accepted a Chair at was also a superb mentor to a number Don’t recall anyone calling for an Strathclyde University in Glasgow. of young statisticians who began their apology! His decision to move to Glasgow may careers in the Division in Sydney in the End of conversation. have been influenced by his two older 1960s. Rupert did spend a year at Cambridge children commencing musical careers Bobbie, Rupert’s wife since 1941, died in 1954, during which Fisher was his in Europe at about that time. There in September, 2006. He is survived by his nominal supervisor. However, because of was also some suggestion that his move daughter Elizabeth (living in Glasgow), the nature of Rupert’s PhD research, from Australia was influenced by strong sons Michael (Munich) and Julian (now he worked with Richard Gregory in the disenchantment with the Australian at Macquarie University in Sydney), and psychology school. Government of the day. three grandchildren. 10 SSAI Newsletter – June 2007 Biometrics by the Beach Member News The International Biometric Society reception that evening. One of the courses Foreman Prize Australasian Region (IBS AR) will hold is on model-based geostatistics, presented Janice Wooton is the eleventh recipient its next biennial conference from Sunday by Peter Diggle and based on the recently of the Ken Foreman award which was 2 December to Thursday 6 December published book by Diggle and Ribeiro. instituted in commemoration of the 2007 at the Novotel Pacific Bay Resort in Full details of all courses will be included contribution made by Mr Ken Foreman Coffs Harbour, NSW. on the conference website as soon as they to the Australian Bureau of Statistics With Biometrics “safeguarding our are finalised. over many years. Janice’s award is in future: our health and our air, land and The conference centre at the resort is recognition of her major contribution water” as its general theme, the program recognised as one of the most attractive will feature invited speakers Jim Zidek to improving the methodology for and workable venues in Australia while census tabular confidentiality, as (University of British Columbia, Canada), the resort offers 4.5-star accommodation well as her contribution to a range Peter Diggle (Lancaster University, UK and facilities set amongst expansive of projects on business surveys and and Johns Hopkins University, USA), landscaped gardens on Charlesworth Bay, microdata access. The award committee Nick Horton (Smith College, USA), with the beach just a short stroll from Allan Donner (University of Western also noted the role she has played in the conference centre. To make the most Ontario, Canada), Kerrie Mengersen the young statisticians section of the of the wonderful location, the program (Queensland University of Technology), includes a break on Tuesday afternoon Statistical Society of Australia and her Russell Millar (University of Auckland), when delegates can take advantage of involvement in organising the 2007 Ross Sparks (CSIRO Mathematical and organised activities including a guided Young Statisticians Conference. Information Sciences) and Ari Verbyla walk in Dorrigo National Park, a kayaking The purpose of the award is to provide (CSIRO Mathematical and Information Sciences and The University of Adelaide). adventure or a visit to a local winery. professional development opportunities in Contributed papers or posters related Please check the conference website the form of overseas travel to a conference to the conference theme or on other www.biometrics.org.au/conf for further or short course. The award is open to biometrical topics are welcome, with information on registration, the program, Methodology Division staff within the abstracts due by 3 August 2007. courses, the venue, accommodation and ABS who would not normally have the On Sunday 2 December, the program travel to Coffs Harbour. We look forward opportunity to undertake a professional will feature two or three short courses, to seeing you there! overseas trip. Applicants are assessed on preceding the conference welcome Melissa Dobbie their technical ability, their contribution to methodological work in the ABS, their contribution to raising the profile of the ABS as a professional organisation; and their potential for further development, Looking for a job? particularly their potential to benefit from For a listing of current statistical the proposed trip to be undertaken. Ken Foreman was the founder of what vacancies in Australia or New Zealand visit: is now the Methodology Division in the ABS. He joined the ABS in 1951 and retired in 1984. He is referred as “the http://www.statsci.org/jobs father of statistical methods in the ABS” (Bill McLennan, 1996). He pioneered Do you have a job to advertise on ABS work on sample design and estimation, development of statistical the website? infrastructure, quality improvement and total survey design, time series analysis, Email a position description to econometric modelling and data analysis. He introduced area sampling for [email protected]. Listing is free! household surveys, generalised statistical processing systems, seasonal adjustment, This service is proudly brought to you by: the survey sampling training courses, and many other achievements which underpin much of the current ABS methodological environment. A ceremony was held recently at ABS Statistical Society of Australia Inc. House in Canberra when Susan Linacre, Deputy Australian Statisician presented the award to Janice. SSAI Newsletter – June 2007 11 CensusAtSchool International Workshop Report to the SSAI by John McKinlay academia, education departments and implemented for unifying international As many will know, Australia is currently statistical societies in France, Singapore, databases and output. Ideas about how facing a severe shortage of mathematical Canada and Australia. the global project might be managed and modellers and statisticians that will The C@S program has a noble and evaluated were also discussed, as were the increasingly inhibit business, industry simple objective: increase statistical literacy mechanisms that might be developed for and government agencies alike. This was among schoolchildren and this will have implementing an international network for recently highlighted by an open letter to the positive, consequent flow-on effects for an students. Significant progress was made Prime Minister signed by 113 international adult population. Anticipated long-term towards achieving these aims and an action and 396 local academics urging action outcomes include a general population plan to further the work was established. be taken on recommendations contained who can think critically about statistics and All delegates rated the workshop as in the recent National Strategic Review of a student population less reserved about valuable, stimulating and productive Mathematical Sciences Research In Australia. undertaking further studies in statistics- in progressing the goal of a truly It was against this backdrop that I had the related areas. The program has proved to international C@S project. The main pleasure of attending a recent international be a successful strategy for introducing outcome was the establishment of an workshop on the CensusAtSchools into school curriculum real data of interest international committee to oversee program. Operating in several different to students. A feature of this approach is further growth in the C@S project in the countries, CensusAtSchools (C@S) is an that statistical concepts and practices are international arena, with some initial work initiative aimed to promote and encourage being introduced and taught in a wide done on developing common questions statistical literacy among primary and range of syllabus, not simply traditional and a single entry point for the project. secondary students. Combining real data mathematical areas, since data that are Countries that do not currently run a about students with stimulating teaching collected may be relevant to a range of C@S project will be offered assistance material, the program has the longer- subject areas. Workshop participants in setting up the program. A second term potential to increase interest and were keen to see this program expand to international meeting has been scheduled participation of tertiary students in the become a truly international project from for March 2008 in Los Angeles. mathematical and statistical streams. which students could access data about From a personal perspective, I was The workshop was held in Melbourne themselves and their fellow students from impressed by the enthusiasm and dedication from March 26 to 29 and was sponsored across the world. of all participants, but particularly the by the Australian Bureau of Statistics The aims of the workshop were to bring delegates involved with teaching children and the SSAI. Brian Pink, the Australian together countries currently operating C@S in schools. Most of us who have been Statistician, opened the conference and programs to establish a plan for each project involved in teaching recognise that the guest presenters included Professor to have an agreed set of internationally difference between an enthusiastic student Neville Davies (UK Royal Statistical comparable data. With each country and an indifferent one is their exposure to Society), Dr Juana Sanchez (Director currently having its own unique set of quality teachers, people who not only teach of the International Statistical Literacy questions, circumstances and imperatives, but inspire. I left the workshop inspired, Project, UCLA, USA), Professor Chris this turned out to be a tricky, but not heartened that not all is lost in lower Wild (University of Auckland, NZ) impossible, undertaking. The meeting was education with input from such dedicated and former Australian Statistician successful in initiating the development of and enthusiastic individuals. Now, does Dennis Trewin. Other participants came a common set of international questions anyone have any ideas about how we can from national statistical organisations, and policies that might have to be fix our universities. Delegates from the CensusAtSchool International Workshop. 12 SSAI Newsletter – June 2007 Branch Reports NEW SOUTH WALES of statistics with the aim to discuss how Hyperspectral images combine spatial and to increase statistical thinking in secondary spectral information to record mixtures of schools. Finally, Anthony Davison, Professor different pure materials across an image. For March 21 saw us hold our first meeting of Statistics at Ecole Polytechnique Federale each pixel in the image, the reflectance at many in access grid rooms linking UTS and de Lausanne, Switzerland will be visiting different wavelengths is measured, producing University of Wollongong. I would like to Sydney towards the end of July and has agreed a spectrum at p wavelengths (p-spectrum). offer a vote of thanks to the people who to do a joint presentation for UNSW and The goal of identifying pure materials and provided us with technical support: Tony SSAI. His areas of interest include statistics mapping their abundances across an image Maher and Corley Kinnane from UTS and of extremes, likelihood asymptotics, bootstrap can be realised using the Iterated Constrained the corresponding person at UoW (apologies and other resampling methods and statistical Endmembers (ICE) algorithm. The ICE for not remembering the name). modeling so I am sure any topic he chooses algorithm fits a convex mixing model by There could be no more auspicious to present on will provide us with much food minimising a goodness of fit measure with a occasion to test out the technology than for for thought. penalty related to the distance between end- our AGM and Lancaster lecture. Although I am looking forward to seeing our members (pure components). the technology was working pretty well (UTS members and visitors at one or more of these Such imaging has applications in fields had trouble understanding people talking at events. Feel free to come up and introduce such as biology, environmental science, UoW but that wasn’t too much of a problem yourself to me or one of the other Council geology, industry and pharmacology. as our guest speaker, Chris Heyde, was members – we would love to meet you. A biological example was given in which speaking at the UTS venue) the link decided Caro Badcock the ICE algorithm can be used to identify to disappear while we were all enjoying pre- anomalies (for example, cancerous cells) in a meeting refreshments! Technical support President NSW Branch tissue sample. worked under extreme pressure, with us all looking over their shoulders to reconnect WESTERN AUSTRALIA A problem often encountered when using us. In addition to the technology challenges the ICE algorithm is that of systematic the seating arrangement at UTS was also noise. This can be introduced to the spectral The seminar on April 17th was presented interesting. Those who joined us had the data via physical causes such as changing by Alex Stuckey. Alex is currently on leave opportunity of sitting in chairs along the sunlight or modeling causes such as from the ABS and is studying towards a PhD back wall, chairs around the desks, a 3-seater nonlinearity in the mixing of end-members. at UWA. Alex’s talk, entitled “Single-index lounge brought in to provide extra seating or To remove such systematic noise, Tony models for panel and spatial data”, was based the floor (lucky it was carpeted)! It was great presented an extension to the ICE algorithm on his PhD studies to date. to see so many people that all these seating that includes the simultaneous modeling options were required and I think added to Alex gave background information on the of spectral baselines. These baselines can the novelty of the event. Not unexpectedly structure of panel and spatial data and also capture the systematic noise and allow it to our speaker handled all these challenges with on semiparametric models, in particular the be removed from the hyperspectral image. aplomb and provided us with an entertaining single-index model. The single-index model Tony proposed a method for reducing the presentation focusing mainly on climate reduces the dimensionality of a nonparametric dimensionality of this augmented problem change. Another pleasing outcome of this modelling problem by creating an index that to allow model fitting. night was the number of young statisticians is a linear combination of the explanatory Tony’s talk was very well received and who joined us and subsequently joined the variables. One then uses nonparametric promoted some good discussion with the techniques to model the response as a function speaker for dinner. A very interesting evening audience members. The audience then retired of the index. Such methods are popular in the all round! with the speaker to a local restaurant for a time-series and regression settings. Alex and Due to Easter and ANZAC Day the NSW Vietnamese banquet. his supervisors wish to extend these methods Branch decided not to hold an April meeting Marty Firth but has still been working very hard on the to panel and spatial data. program of events for 2007. I would like to The methods were illustrated using SOUTH AUSTRALIA remind everyone to check the website for simulated spatial data and an economic panel the updated program. Our June speaker will dataset. Potential applications in the field of be Adrian Baddeley with the venue probably climatology were also discussed. being CSIRO or Macquarie University, both The theme for the May branch meeting Analysing administrative by product data at North Ryde. Details of Adrian’s talk will was “An Evening with Young Statisticians”. – is it worth the effort? have been finalized while this article was Those attending were treated to a very Following the Annual General Meeting going to press. interesting presentation by Tony Traylen, a in March, the outgoing President of the SA At this stage we are looking to have recent graduate now working at the CSIRO. Branch, Margaret Swincer, gave a talk about three events in July. On July 17-18 Kerrie Tony was awarded the SSAI WA branch her experiences working with administrative Mengersen will be running her Bayesian honours scholarship last year. Tony’s talk, by product data. Sources of this type of data are workshop. July 25 will see us holding an Estimating spectral baselines in hyperspectral varied and include taxation, unemployment, inaugural afternoon workshop aimed at images, first gave background on hyperspectral pensions, health systems and compensation. secondary school mathematics teachers and imaging and then focused on the problem of The data are collected to monitor and manage university lecturers interested in the teaching removing systematic noise from these images. processes and to comply with legislation. SSAI Newsletter – June 2007 13 Branch Reports Continued Working with administrative by product “clients” are the program participants. Much of analysing surveys, as well as opportunities to data is a challenge for the statistician. Typically the statisticians’ work focuses upon evaluating delve into qualitative research. the data are contained in multiple databases the efficacy of these programs in achieving The evening concluded with a few bottles coming from different sources, so there is a job placement for their clients. of wine, good conversation and a Turkish need to first understand the complex data One of the key tools at DEWR is a Star banquet for 15 people at a local restaurant. structure. The issue of changing technology rating system which is used to compare Terry Neeman is a problem, as data are often collected the effectiveness of government-supported over time. The integrity of the data is often agencies in finding employment for their Talk by Matt Wand questionable and data cleaning is a major task. clients. They use logistic regression to Due to the size of the data sets, statistical At the monthly meeting of the Canberra model the expected probability of a client programming is required in a language Branch of the Statistical Society on Tuesday obtaining employment conditional on client which can handle millions of records, such 20 February 2007 Professor Matt Wand of characteristics, local area job availability and as SAS, in order to extract the relevant data. the University of Wollongong (and recently other social indices. These expected outcomes Statisticians working with this type of data of the University of New South Wales) are compared with actual outcomes, and may have to deal with professional isolation gave a talk titled “Semiparametric regression agencies are awarded stars from one to and a lack of recognition of the need for and the computer science interface”. The professional development. However this is five (five being the best) based upon their presentation was followed by a free barbeque particularly important for statisticians in this ranking relative to their expected outcomes. dinner to welcome new members of the field, as there is currently a lack of appropriate Future funding of agencies may depend Statistical Society. statistical methodology for dealing with this upon these ratings. Matt began his talk by summarising the type of data. Data linkage has a major role to But it’s not only about the numbers. Sepi field of semiparametric regression, with an play, although there is the question of whether emphasised that qualitative research plays illustration involving longitudinal data on the all data sources actually go together. an important role in the evaluation process. spinal bone mineral densities of individuals The talk concluded with a lively discussion Sometimes the numbers alone do not make within certain racial groups. Referring to about how statisticians can get across the sense, and direct client contact is needed his book “Semiparametric Regression” (co- importance of complex results to the relevant to put a human face on the data. Sepi has authored with D. Ruppert and R. J. Carroll), people, and to make them understand that done a good share of this qualitative research, he discussed the generalised linear mixed statistics is more than just cross tabulations. traveling around Australia to interview clients model, its ability to readily incorporate longitudinal and spatial effects, and its ease Lisa Yelland and the agencies who serve them. This research, or “case studies” as she called them, of implementation in SAS, R, S-Plus and WinBUGS. Matt then related various facets of CANBERRA is an important supplement to the data which have also enhanced her understanding of the statistics with their counterparts in computer data she analyses. science (e.g. missing data methods with belief networks). This led on to a discussion of DEWR also aims to evaluate the net impact Statistics at the Department of data mining, which he has been teaching of back-to-work government initiatives Employment and Workplace Relations - with an emphasis on business applications by comparing employment placement of (DEWR). - for a number of years as part of the UNSW persons participating in these programs with Master of Statistics course. The recent recruitment drive for employment placement in matched cohorts. statisticians and other quantitative analysts Subject to the limitations of cohort studies, Data mining is concerned with the to the Department of Employment and the results of these analyses indicate that the opportunistic discovery of previously Workplace Relations (DEWR) has brought programs Customised Assistance, Job Search unknown information from the data in data a higher level of statistical thinking in the Training, Work for the Dole and Mutual warehouses, with a focus on interpretation and support for decisions. There is evaluation of back-to-work programs and Obligation have had a net positive impact on other employment and education initiatives. much opportunity within data mining their clients gaining employment. Sepi Helali, senior statistician at DEWR, was to apply statistical methods, in particular our 1 May featured speaker. She discussed Another recent initiative DEWR semiparametric regression. In turn there are some of the current statistical work as well as has undertaken is a five-year longitudinal many opportunities for computer science some work planned for the future. pathways survey to track cohorts of people in semiparametric regression, for example on income support (including recipients Statisticians at DEWR are mostly in relation to the classification problem, who leave income support) and their job employed in the research and evaluation wherein training data are used to guide the experiences over time. Data will be collected group. This group is responsible for the classification of future objects into two or evaluation of government programs and at six monthly intervals. more classes based on observed predictors. government supported agencies who work Sepi emphasised that DEWR is a great A case in point is the problem of spam, to achieve greater workforce participation. place for statisticians to work. The available namely: How do we teach our computers There are programs tailored to a wide range data are rich and varied. DEWR statisticians to figure out which e-mail is spam and of demographic groups: from young people have access to administrative data, survey data which is legitimate? To answer this question, just entering the workforce, to persons with and program evaluation data. Moreover, there Matt referred to a recent study of his where disabilities, persons on long-term income are many opportunities to apply statistical he examined a large number of candidate support and mature age job-seekers. DEWR techniques in modelling data, designing and predictors, for example the frequency of 14 SSAI Newsletter – June 2007 Branch Reports Continued certain characters and sequences of letters. Professor Kerrie Mengersen. Among the The aim of this study was to come up with a highlights were the Belz lecture, presented classificatory mechanism which has all of the by Professor Rob Hyndman, and a talk by following: parsimony, interpretability, speed, Professor Clive Granger, 2003 Nobel Prize and, of course, accuracy in classification. Laureate in Economics. The Branch also Such a mechanism can be provided by partially supported three Victorian young Laplace approximations of solutions to statisticians to attend the Young Statisticians suitable logistic models with both linear conference in Canberra in April 2007. Brian terms and spline terms. thanked the Council for a very successful One of these models is called the KOW year, especially the efforts of Ann Maharaj as Algorithm (after Kauermann, Ormerod and Secretary and Geoff Laslett as Treasurer. Wand). This model determines the ‘best’ linear and non-linear components via Rao All the activities of the branch are listed Geoff Laslett and Geoff Bruton (right) statistics and cyclically adds them into the on the Branch website, which is regularly Photo: Brian Phillips model according to the marginal Akaike updated. See http://www.statsoc.org.au/ Information Criteron (mAIC), until the Branches/VIC/. After noting that the Branch mAIC fails to decrease substantially. Matt was in a healthy financial state, the Treasurer Developing probabilistic illustrated the KOW algorithm by applying pointed out that 2006 was the first time in statistical thinking it to some contraceptive method choice data many years that the Branch had not financially The speaker after the Annual General and then to some spam data. In both cases supported any ‘Statistical Events’(the ‘About’ Meeting and Service Award presentation the algorithm quickly came up with a useful page of the Branch website gives examples of was Professor Helen MacGillivray of the Queensland University of Technology. Helen and parsimonious model. He then compared events supported in the past). He implored currently holds a Senior Carrick Fellowship the algorithm with some other methods members to put cases of worthy events (those using simulated datasets and discussed a on the theme “The teaching and assessment that enhance the profile of statistics in the number of variants. of statistical thinking within and across community, business, industry or academia) disciplines”. Borek Puza to the Branch Council for consideration. VICTORIA The 2007 Branch Council was elected. It consists of: SSAVIC Listserver President: Mervyn Silvapulle A list of members of the Victorian Branch Past President: Brian Phillips of SSAI has been created on a server at Acting Secretary: Ann Maharaj Monash University. Members can now post Treasurer and SSAI Newsletter Correspondent: messages which will be received by all on the Geoff Laslett list. If you need to advertise an event relevant to the society’s interests, such as a seminar, Council Members: Muhammad Akram, Lucy workshop or job vacancy, you are welcome to Busija, Kym Butler, Kris Jamsen, Kay Lipson, Professor Helen MacGillivray post to the list. The list address is: Debra Partington Photo: Brian Phillips [email protected]. The President thanked outgoing Branch If you are not a member and wish to add Councillors Tristan Barnett, Ian Clark and Helen’s talk was titled “Developing your name to the list or if you are on the list Derchieh Hung for excellent service over the probabilistic statistical thinking”. Her slides and wish to have your name removed from last few years. had the sub-title “Using Data, Student it, email: Experiences and Collaboration in Developing [email protected]. SSAI Service Award Probabilistic Reasoning at The Introductory Ann Maharaj Tertiary Level.” After the Annual General Meeting, Geoff Helen drew on her extensive experience Bruton was presented with an SSAI Service Annual General Meeting to explain how the focus of probabilistic Award by Geoff Laslett. Geoff Bruton joined understanding has been mostly on its role The March meeting of the Victorian the Society in 1970, and had two stints as a Branch was held at the Swinburne University in data analysis and inferential reasoning. Victorian Branch Councillor, completing 11 of Technology on March 26 and was attended She explained how, in recent years, statistical years in total. He was Branch Secretary from by about 30 people. It commenced with education has been data-driven, stating that the Annual General Meeting. The outgoing 1995 to 1999, and Central Council Secretary the holistic “reform movement” has been President, Brian Phillips, gave a summary from 2000 to 2002. Geoff has now semi- influencing courses for “service”, mathematics/ of the activities in 2006. These included retired to Gallery 141 in Red Hill, well worth statistics majors and schools. She said that it is seven monthly meetings held at a number of a visit if you are touring the Mornington time to apply those principles to probabilistic different locations (Swinburne, Monash and Peninsula. He still teaches statistics part-time, literacy and reasoning. Melbourne Universities) plus a very successful partly to support his hungry (and, hopefully, She described two introductory student workshop on Bayesian methods, given by grateful) herd of very cute alpacas! contexts in which she teaches: SSAI Newsletter – June 2007 15 Branch Reports Continued (a) Introductory data analysis courses, processes, and to consolidate and synthesize Queensland Department of Primary Industries which are compulsory “service courses” in fundamental mathematical skills. and Fisheries (DPI&F) at Toowoomba science, engineering and mathematics, as Following the talk there was a healthy joined with Brisbane-based statisticians for well as elective for others. Here “purposeful discussion both formally in the lecture room, an enjoyable and stimulating evening. The probability” (that is, only what is needed) is then less formally at a nearby restaurant. evening began with guest speaker Dr Toshio included. Ohnishi speaking on ‘Estimating a common Brian Phillips (b) Introductory probability and slope (of multiple strata) in the Tweedie generalized linear model using a conjugate distributional modelling, which are QUEENSLAND compulsory in all mathematics/statistics prior’. Dr Ohnishi is an assistant professor with majors, where more extensive probability is April Meeting taught. the Institute of Statistical Mathematics, a The AGM of the Queensland Branch of Japanese governmental institute in Tokyo. He She said that both types of courses the Statistical Society of Australia was held on is also an assistant professor of the Graduate emphasize doing and communicating and 3 April Gardens Point Campus, Queensland School for Advanced Studies (SOKENDAI), her classes are zone free of dice, coins, cards University of Technology. Kanagawa, Japan. He is now visiting the and balls in boxes. Her approach is to use University of Southern Queensland until July purposeful and minimalist probability in Following on at the ordinary meeting, under the sponsorship of SOKENDAI. introductory data analysis. This involves using Professor Annette Dobson spoke on examples from “field” research, introducing p- ‘Missing Data in Longitudinal Studies: In his talk, Dr Ohnishi discussed the values through Chi-squared tests after a study experience from the Australian Longitudinal Bayesian analysis of the generalized linear of Exploratory Data Analysis, but before Study on Women’s Health’. The meeting model based on the Tweedie distribution. introducing any complications of sampling was well attended and over 20 people were Tweedie distributions are a family of distributions. She claimed that interpreting in attendance. Annette described how distributions in the exponential family with “smallness” of chance is the big conceptual missing data can be a problem in large scale variance proportional to some power of the step. For confidence intervals she argues that population surveys. Ignoring missing data by mean. (For example, gamma distributions the biggest hurdle is not probabilistic – rather using a complete case analysis can produce have variance proportional to the square of it is understanding that the interval is for biased results. Biases occur then participants the mean.) In particular, he focused on a estimating a parameter, not for “most of ” the providing complete data are systematically special subset of the Tweedie distributions which are continuous on the positive real axis, individual values. different from those with missing data. Longitudinal studies are especially susceptible plus a point mass at zero. Helen gave examples from exercises and to such bias as missing data accumulate over Dr Ohnishi considered the multiple strata projects used in her courses. She emphasised time due to item non-response, wave non- problem, which has stratum-dependent the use of group project work and has response and participant drop-out. intercepts and a common slope. Assuming a continually modified her course over the years. prior density on the intercept only, a density Conditional probability and Markov chains There is a growing body of literature in the location family of distributions is are important concepts in her courses and she and software for dealing with missing data, proved to be conjugate for the model of gave many examples her students used in their including weighted estimated equations, interest when used with a logarithm link projects. These involved collecting data and multiple imputation, likelihood-based function (the most common choice for these investigating assumptions in situations chosen methods and fully Bayesian models. While multiple imputation procedures are now models). He discussed the conjugate analysis by the students. Examples included: the AFL becoming widely available and practical for the intercepts and derived the optimal grand final, queuing at Medicare, pedestrian (especially using chained regression methods), estimator. He also obtained the optimal traffic in a mall and occurrences of “Harry” in there is much less work on the adaptation of estimating function of the slope under a “Harry Potter and the Goblet of Fire”. these methods for longitudinal data. criterion based on that of Godamba and Kale Helen presented some quantitative data (1991) extended to the Bayesian framework. In the Australian Longitudinal Study illustrating the success of her methods. Since on Women’s health a particular problem is Finally, he revealed an interesting relation the introduction of group projects in 1999, women’s reluctance to report their weight; between the optimal estimator and the there had been increased engagement. When yet weight is an important risk factor for optimal estimating function. data-based class activities and preliminaries chronic conditions. This talk illustrated As the evening progressed, appetites were introduced in 2003, there had been a the use of several methods, including fully were (eventually) whetted and conversation further increase in engagement and student Bayesian models, for handling missing data in stimulated by a late evening at a local approval. And there was a real buzz and a longitudinal analysis of the effects of weight restaurant. high student approval after the 2006 tutorial and weight gain on the incidence of diabetes. group exercises were used. She claimed that While in Australia, Dr Ohnishi plans to the course has been the best predictor of Miranda Mortlock eat a lot of steak, improve his tennis skills, and performance in later courses. She gave a work with colleagues in Toowoomba at USQ reason for the success of her programs: the May Meeting in Toowoomba and Department of Primary Industries and students have gained generic skills, learning On 8th May, the Queensland Branch of Fisheries to apply the methodology espoused how to unpack the known, to analyse, extend the Statistical Society met at the University of in his interesting talk. and synthesize, to tackle real problems, to link Southern Queensland (USQ), Toowoomba. Peter Dunn with data, to carry out real investigations and A strong contingent from USQ and the University of Southern Queensland 16 SSAI Newsletter – June 2007