statistical society of australia incorporated newsletter

February 2004 Number 106

Registered by Australia Post Publication No. NBH3503 issn 0314-6820 Inaugural Queensland Statistics Conference

The inaugural Queensland Statistics same dataset throughout the session. Conference took place from 1-3 October Procedures for modelling with the 2003. The 2nd of October saw the first data were covered, followed by useful day of proceedings, and was a cold and approaches for plotting the main effects blustery day. However, the welcome by and interactions of the models. Lastly, conference organisers in the tea room of Ross demonstrated code to examine the the University of Southern Queensland validity of any models derived, and the Department of Mathematics and significance of any random and fixed Computing was warm enough for effects terms in the model. the coldest of climates. With tea and Peter Baker discussed the value of the coffee being served during registration, R packages, Emacks Speaks Statistics conversation centred on the R workshop (ESS), Bioconductor, BugsR and bqtl/qtl. jointly presented by John Maindonald Each package has its own utility. ESS is (ANU), Ross Darnell (UQ), Peter Baker designed to improve the user interface (CSIRO) and Peter Dunn (USQ) the of S and R. Bioconductor is a work in previous day. A rousing success by the progress of an open source software sounds of it, I was disappointed I was package for the analysis of genomic unable to attend. data. BugsR (bugs.R) is a package to use Attendees were taken through an WinBugs in R, though it is not available introduction to R with John Maindonald, at CRAN. The last two packages bqtl starting by familiarising themselves and qtl are used in analysis of data from with the R environment, inputting data studies in population genetics. and basic plotting methods among The last session of the day was other things. Things moved pretty presented by Peter Dunn on “Sweave” quickly and soon complicated plotting which is a framework for mixing R techniques that get the most out of and LATEX. This allows graphics and the graphical capabilities in R were models produced in R to be used as demonstrated. embedded code in LATEX documents. Ross Darnell led the second session of This enables graphics to be updated the workshop, this one on Non-Linear very easily within these documents Mixed Models in R. The focus of the without the need to reproduce and then session was the “lme” package in R, insert graphics. designed specifically for this type of Dr Baker solves the problem of arriving Many of the presentations given modelling. Attendees looked at the late to meetings on Thursday, the first official day of proceedings, examined various statistical applications to the natural In this issue sciences, including ecology, agriculture and medical studies. Alex Creagh President’s Corner ...... 3 ISI 2005 ...... 14 from Griffith University discussed the Letters to the Editor ...... 5 BioinfoSummer ...... 17 applicability of case control designs Editorial ...... 6 Branch Reports ...... 18 to the study of natural disturbance Section News ...... 11 Conferences ...... 24 events in rainforest ecosystems. There 2003 Ken Foreman Lecture ...... 13 Report on Biometrics 2003 ...... 26 were also several presentations on novel statistical methods and efforts to Inaugural Queensland Statistics Conference Continued improve established techniques, such Friday saw many more presentations of the conference, a prize of $250 from as the presentation by Olena Kravchuk from students, researchers and the Queensland branch of the SSAI from the University of Queensland on a applied statisticians including the and a $50 book voucher from Pearson new rank test of location equivalent to two invited speakers Bob Murison Education to the runner up. Ben the Wilcoxon two sample test, suitable and John Maindonald. Bob Murison Stewart-Koster from Griffith University for the hyperbolic secant distribution. from the University of New England won the $250 for his presentation on Also included in Thursday’s also discussed SAMM and ASREML, “Statistical modelling and the problem programme were presentations by identifying examples where this S- of scale in ecology” and Rodney Ellis invited speakers Dave Butler and Plus module is most valuable. John from the University of Queensland won Kerrie Mengerson. Dave demonstrated Maindonald gave a presentation the book voucher for his presentation the value and applicability of SAMM, called “Predictive validation issues titled “The use of correspondence an S-Plus module for mixed models in data mining” where he discussed analysis applied to marine data”. using restricted maximum likelihood approaches (both theoretical and All in all the first Queensland (REML), with spatio-temporal empirical) for assessing the predictive Statistics Conference was a great success data from a white clover breeding accuracy of models. John is the senior with speakers from around Australia program. However, he left most of author of the book “Data Analysis and contributing to the program. Having the examples to his colleague Alison Graphics Using R” recently published the conference in a regional centre was Kelly, who spoke right after him. Dave by Cambridge University Press. a great idea. It certainly opened my eyes is a consulting statistician with the Another highlight was from Daniel to the possibilities of future statistical Queensland Department of Primary Burrell, a student from the University of Industries involved with crop genetic Southern Queensland, who presented a research beyond the traditional centres improvement programs. Kerrie paper on a new goodness of fit statistic in Brisbane and the other capitals. The Mengerson is a Professor of Statistics for generalised linear models. success of the conference was due to at the University of Newcastle. Kerrie’s The conference concluded with the fantastic efforts put in by Peter presentation focussed on a key research closing comments from Bronwyn Dunn (USQ) in particular and also the interest, Bayesian modelling. These Harch (CSIRO), the president of the assistance provided by Ross Darnell methods were presented in the context Queensland Branch of the Statistical (UQ) – a big thank you to them! of case control and cohort studies into Society of Australia, who awarded the Ben Stewart-Koster the effects of passive smoke. prize for the best student presentation Honours Student – Griffith University IBC 2004 XXIInd International Biometric Conference in parallel with

ASC 2004 11-16 July 2004 Cairns Convention Centre, Australian Statistical Conference Cairns Queensland Australia

w.ozaccom ww .com.au/cairns2004

Early Bird Registration Closes 1 March!

Statistical Society of Australia Inc.

2 SSAI Newsletter – February 2004 President’s Corner

ASC 2004 SSAI Membership 1994 – 2003 ASC 2004 will be held in Cairns Full Members Student/Retired Members at Large from 11th to 16th July in parallel 1,000 with IBC 2004 and a broad-ranging program has been organised 900 (go to the conference web-site http://www.ozaccom.com.au/ 800

cairns2004/ for up-to-date 700 information). Over 500 abstracts

had been submitted for oral 600 and/or poster presentations when submissions closed early in 500 December. It is great to see such strong interest in the conference. Early bird registration closes 400 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 on the 1st March so there is not a lot of time if you want to take advantage of the discount available. I have already registered Membership and look forward to seeing you there in Cairns. It is great to see that SSAI membership is increasing again Daryl Daley Festschrift although we have some way to go before we get back to the levels of the early nineties. Congratulations to the branches and Daryl Daley will turn 65 early in April and SSAI/NZSA will everyone else who has endeavoured to ‘turn things around’. recognise Daryl’s contribution to probability and statistics in Australia and internationally by publishing a festschrift issue of the Australian and New Zealand Journal of Statistics. Peter ASC/NZSA 2006 Taylor and Philip Pollett have arranged papers from an eminent The 2006 conference will be held in Auckland and be list of contributors and this is a measure of Daryl’s standing conducted in co-operation with the NZSA. A memorandum worldwide. A special session in Daryl’s honour will be held at of understanding between the two societies has been agreed, a ASC 2004 and a list of the festschrift papers can be found at the conference organiser has been appointed and detailed planning conference web-site (see above). Details of the ASC 2004 session will get underway shortly. There is plenty of scope for members will also be included there as soon as they are available. to participate in a variety of roles (either on the organising committee or on the program committee). Here is your chance to ANZJS make a contribution and to have some influence on the outcome. Presentations about the possibility of ANZJS becoming Simply contact me directly or anyone else on the executive or electronic-only have been held in Canberra, Adelaide, Perth branch council. and Brisbane. A session will be held in Sydney late in February. Neville Bartlett Thank you very much to all of those who attended and Email: [email protected] contributed comments, suggestions and views. The SSAI/ NZSA group considering the matter have had a telephone conference to discuss progress and would welcome any further suggestions. A brief summary of the issues and progress towards a recommendation is provided elsewhere in this newsletter. Volunteers It has been claimed that the Society is run by an elite group of people. Having been a member of the Victorian branch council (in the past) and central council (at present) I can assure you that there is a great diversity of people involved and that the eliteness is purely based on willingness to contribute time and effort to the society’s activities. Enthusiastic (new or old) blood is always welcome. There are a few roles that require an apprenticeship www.ozaccom.co (in some form) but most do not. I have found that my roles in m.au/cairns2004 the society have provided a valuable way to interact with other members of the profession. It is much better generally to have a Sinking mystery large number of people involved in society activities as this helps to In 1980 the 90,000 tonne bulk ore tanker MV Derbyshire sank with build a healthy and vibrant organisation. As you will see from the all hands in a typhoon, south of Japan – the largest British ship ever lost at sea. points below there are plenty of ways to contribute. How about it! Analysis of the wreck 4 kms beneath the Pacific revealed it had sunk when the forward cargo hatch failed – and the crew was blamed. ISI 2005 But statisticians were able to show the loss was really due to gradual The International Statistical Institute will be holding a very leakage through forward ventilation pipes forcing the bows lower in the water, until pounding waves finally burst the hatch. large conference in Sydney from 5th to 12th April 2005 (see The crew were exonerated posthumously, thanks to statistics solving http://www.tourhosts.com.au/isi2005/). SSAI is assisting with a problem that had puzzled investigators for 20 years. the organisation of this conference and we are seeking volunteers Statistics: a job for professionals to assist with the contributed papers that will be submitted later www.statsoc.org.au/PublicAwareness in 2004. Some effort is required in the next few weeks but the bulk of the work will occur around November of this year.

SSAI Newsletter – February 2004 3 Should ANZJS become electronic-only?

Introduction clarity of material could be the only criteria download the files. The permanence of This note briefly describes the main for acceptance of papers with extremely hardcopy needs to be matched by some differences between publishing with long or very terse papers being rejected. In off-line form of the electronic version (such hardcopy and electronic-only publishing principle, papers could be as long as authors as CD-ROM or DVD) being made available along with an update of feedback from wished but extreme length may reduce the to members. Numerous other suggestions members and the steps that are being taken succinctness of the exposition. Papers have been made. towards a recommendation. A series of would be published as soon as they were Institutional subscribers presentations to the SSAI branches is nearly ready and not have to wait for a suitable slot complete and these sessions have resulted in the printing schedule. Cost savings (due Most institutions recognise that the in a wide variety of views being expressed to the removal of printing) could be used majority of access to journals is done along with many useful suggestions. to provide enhanced electronic capability. electronically but they are quite reluctant A trans-Tasman group of four people Members requiring personal copies (albeit to give up hardcopy just yet. This is partly (Murray Jorgensen and Russell Millar electronic) of papers could be supplied with because a major publishing agent went from NZSA with Chris Lloyd and Neville a CD-ROM/DVD version. bankrupt and left subscribing institutions Bartlett from SSAI) have been charged with with hardcopy as their only form of examining the various options and putting Disadvantages of access until alternatives were found (and a recommendation to both societies. A poll electronic-only publishing purchased). A survey of institutions that of members may be used as part of this Should the ANZJS become electronic- subscribe to ANZJS is underway but only process. only, some members feel that the perception 10% have responded so far and 40% (17 of the journal will suffer in the eyes of the out of 43) of these are not prepared to Current situation profession generally. No mainstream re-subscribe to the journal in an electronic- only form. A higher response rate will be The ANZJS is available in hardcopy statistics journal has successfully made needed before any firm conclusions will be and also in an electronic version. Access the transition to be electronic-only and it made on this front. to the electronic version is available to all is believed that concern about this point SSAI and NZSA members by going to the is holding existing journals back from the It should be noted that SSAI and NZSA Stats Web site (http://www.statsjournal abandonment of hardcopy. Many journals jointly hold the copyright to the ANZJS. In sweb.com), click on ‘Browse Journals & are undergoing the step of adding electronic the unlikely event that our publisher goes Societies’ and logon under Australian & access as well as maintaining hardcopy bankrupt, is purchased by another business New Zealand Journal of Statistics. When versions. Hardcopy is seen as being or ceases to operate then we have control prompted for a password enter ‘analysing’ dependable (always there) and a drop in over who can publish the journal. Alas, (without the delimiters). The pdf files that readership is feared if the regular arrival if we do not have archival copy then this can be viewed here cover the years 1998 of the printed edition no longer occurs. may provide some practical difficulties for to the present issue. A search capability Concern about electronic archiving and a while. is available. This access is quite basic and access for some members are also issues. Aspects that need to be changed does not have sufficient functionality to be the only form available to members. Feedback from members Archiving of electronic copy is one In addition to reinforcing the main area that needs to be tackled regardless of Advantages of points in favour of or against electronic- whether the journal becomes electronic- electronic-only publishing only publishing a number of other points only (see comment in the previous section). Use of colour, sophisticated graphics, were made. The printing characteristics of Improved electronic access for members demonstrations and data sets would some web-based applications have proven who are not covered by organizations that are institutional subscribers is another become available with electronic-only to be a source of frustration and will affect worthwhile aspect that is being considered. publication. There would be hyperlinks acceptance of any new electronic version The currently available mechanism via to references and no need to publish four of the journal. Printing pdf files does not StatsWeb (see above) is poorly known to issues of 128 pages a year. Quality and pose any difficulty apart from the time to members and has limited functionality.

Highlights of the forthcoming September Issue 46(1) of the Next steps in the decision Australian and New Zealand Journal of Statistics is a making process 192 page festschrift in honour of Daryl Daley. The review committee are examining Contributors to this special issue are Joe Gani, Peter Whittle, Tony Pakes, another Blackwell journal (Journal of Stephan Böhm, Lothar Heinrich, Volker Schmidt, Chris Heyde, Geoff Laslett, Risk and Insurance) that has an enhanced J. Paige Eveson, Tom Polacheck, Dietrich Stoyan, Helga Stoyan, Gunter Döge, electronic version. This is being considered J.F.C. Kingman, Stephen Cornell, Valerie Isham, Peter Hall, Nader Tajvidi, as a potential model for ANZJS but Phil Pollett, Hanjun Zhang, Naoto Miyoshi, Tomasz Rolski, David Vere-Jones, the feedback so far is not all positive. Frederic Paik Schoenberg, Aleksandras Baltrūnas, Claudia Klüppelberg, R.A. Institutional subscriber reaction will have Vesilo, M.R. Leadbetter, G.V. Spaniolo, Frank Ball, Robin Milne. This impressive a strong influence on what options are list is testament to the esteem in which Daryl is held by the international applied feasible and further responses will be probability community. sought. Once a clearer set of practical This special issue was edited by Peter Taylor and Phil Pollett. I would like to options are available along with robust thank them for their special efforts in producing such an outstanding issue. costing then the review team will consider what recommendations to make. Chris Lloyd Neville Bartlett

4 SSAI Newsletter – February 2004 Letters to the Editor

Dear Editors Dear Editors irrelevant to the making of inferences. On I have been a member of the SSAI for In the November 2003 issue of the SSAI the other hand the training of researchers many years. I was against accreditation newsletter Murray Cameron commented in statistical concepts and thinking has when its introduction was being discussed, that in relation to traditional biometric increased their understanding of the and I am still against it. Please be advised work “Better training of researchers power of sound statistical practice. They that I do not want any of my membership and better software have reduced the realize that it enables them to design fees to be used to support the accreditation demand for statistical innovation”. Our their experiments more effectively and to process or any associated advertising. experience is the exact opposite. extract more information from their data. I realize that the debates about It is true that improvements to statistical They still require help from practising accreditation will be endless but I still wish software continue to make more and statisticians, but the help they require is to add my two bits worth. First of all, in my more powerful and sophisticated often at a more advanced level than it thesaurus “accredited” is not a synonym of methods available on researchers’ would have been in the past. “competent” and “non-accredited” is not a desktops. However many researchers Why do statisticians continue to sell synonym of “incompetent”. An interesting do not have the understanding to use themselves short? To quote John Nelder situation arises when an accredited these methods effectively, and if they (The Statistician, 1999) “the public image statistician commences work in a new field do have this understanding it is not of statistics is poor and may be getting e.g. moving from engineering applications practicable for them to spend the time worse. Almost nobody knows what to agricultural applications. I believe that necessary to maintain it. Methodologies statisticians do, and we in turn have been which researchers have difficulty with even an accredited statistician will initially remarkably ineffective in explaining to include many of the modern approaches be prone to making possibly naive errors non-statisticians what we are good at”. to the analysis of data with multiple until he or she has gained some experience The comment by Murray is just another sources of variation, frailty models, in the new field. How long will it take for example of why statisticians will lose generalized additive mixed models, etc. the accredited statistician to gain sufficient further influence and, if it continues, will Many researchers seem to have a poor experience to be a competent agricultural ultimately not survive! statistician? I believe there can be times understanding of how to use regression and situations when an accredited splines appropriately. Yours sincerely statistician is not fully competent and may Much of what has been taught to non- Jeff Wood and Ross Cunningham not even be aware of it! It seems logical statisticians is not very useful and often Australian National University to me that the supporters of accreditation should therefore insist on accreditation for only the extra fees paid by accredited rah-rah for the statistical profession my particular fields rather than some sort of members and any corporate donations overall impressions were quite favourable. generic accreditation. Will they? Of course provided explicitly for this purpose. With It did not enlighten me further with I believe that they won’t but I also believe regard to the Campaign I confess that the regard to the Challenger disaster but this is an important issue because, to my Pryor cartoon on p.7 of Newsletter No. this is perhaps understandable given the horror, the SSAI is portraying accreditation 104 leaves me cold and the caption below booklet’s purpose. I thought the “Census as some sort of certificate of competence. it does not get me excited (of course these discrimination” example was unwise, How many members support are very personal reactions). The caption especially given that the internment accreditation? How many members intend does not provide enough information during a major world war may even to seek accreditation? How many members about the calculation of risk of failure have saved some lives despite the are accredited? Can we please have some for the ill-fated Space Shuttle Challenger obvious negatives. Can a suitable peace up-to-date answers to these questions? We mission so it is difficult for me to make time example be found rather than risk seem to be approaching a two tier society an informed comment. Many thoughts raising very complex moral issues in apparently intent on providing maximum passed through my mind when I first read which avoiding ethnic discrimination benefits ONLY to accredited members. For the caption – e.g. the joke that 49.7% of may have actually led to more death and example I am offended that accredited all statistics are made up and the rest are suffering? With regard to “Patching the members may be entitled to reduced wrong; Han Solo’s remark to “never tell missing holes” the statistician’s logic was, costs for attending workshops etc. Let me the odds” (would even the relatively in fact, potentially flawed but we are not the accredited members be satisfied with high risk of failure as estimated by the given enough information to vindicate the the improved employment prospects they statisticians been enough to persuade statistician’s suggestion. Besides this, the presumably believe accrue to them by NASA management to abandon the statistician’s suggestion could have been being accredited. I read that a fund has mission, given that they went ahead with offered by any one with problem solving been established to promote accreditation the launch even though the air temperature skills, it did not require any statistical through a combination of corporate was more than 20 degrees Fahrenheit training whatsoever, and I believe that donations and surpluses from Continuing cooler than the recommended minimum many statisticians would not have even Professional Development workshops. If which they had apparently been advised); thought of it. I suspect that the “Smoothing the corporate donors know that they are and in any case did the calculated risk of the Internet” example was simply either a supporting accreditation then I have no failure pinpoint the infamous O rings as a dumb analysis or dumb reporting or objections, but I again find it offensive if likely cause of failure? both, and that Internet performance must workshop surpluses are being used to Anyway, after seeing the Pryor eventually reach a peak before declining support or promote accreditation. cartoon I decided I should track down rapidly – perhaps much depends on what I also hope that the Public Awareness the corresponding booklet of Success is meant by “full to capacity”. I think Campaign to support professionalism and Disaster stories which I duly found readers who only read the first paragraph and accreditation is being funded using on the SSAI web page. As a bit of rah- of the Internet example would be shaking

SSAI Newsletter – February 2004 5 Letters to the Editor Continued

their heads in disbelief. If the booklet was resulted from some youthful exuberance promoting the employment of professional or be an example of statisticians having fun statisticians on statistical tasks then my with numbers. Putting these possibilities overall satisfaction rating would be high. aside, my own thoughts are that we are However, its major emphasis is on the really dealing with a population of 25, or PO Box 4085, employment of accredited statisticians and perhaps with a nonrandom subset of all Ainslie ACT 2602 this I again find offensive. The overall tone of the people who ate the lunch. In any Phone/Fax (02) 6249 8266 is that only accredited statisticians can help case no confidence limits should have Email: [email protected] employers, with a hint that non-accredited been calculated or reported. If I am correct statisticians may be unprofessional or then what are we to think of the slogan Society Web Page incompetent. And why the emphasis on “Statistics: a job for professionals” when http://www.statsoc.org.au Government departments instead of all the SSAI cannot even clean up its own employers? I work for a Government Newsletter? department. If I continue to reject Please, please be assured that this is Editors accreditation then it appears that I am not a personal attack on any individuals. in the situation of belonging to a society There have been other instances in the Alice Richardson, School of Mathematics which is systematically attempting to past when I could have questioned and Statistics, University of Canberra, take away my livelihood. I feel my long- the validity of certain statements in PO Box 1, Belconnen ACT 2616 Email: term future may be threatened so with the society’s Newsletter. I am simply [email protected] mixed feelings of sad reluctance and grim against accreditation of statisticians here Fax: (02) 6201 2683 determination I offer the following: in Australia and am merely intent on Michael Adena, Covance Pty Ltd I believe we humble members are forcefully expressing my view. If we really PO Box 4064, Ainslie, ACT 2602 encouraged to report published examples must persist with accreditation my own Email: [email protected] of bad statistical practice to the Society for preference would be that all publicity, Phone: (02) 6274 9704 investigation (presumably by an accredited booklets etc. push ALL employers statistician!) so I wish to report the Society’s for employment of PROFESSIONAL Correspondence own newsletter. An example occurs in the statisticians and that accreditation receive Continuing Professional Development much less emphasis, something like an Please direct all editorial correspondence optional extra. Does the SSAI exist for to Alice Richardson. section (!) on p.17 of Newsletter No. 104. In particular I refer to the 95% confidence all those with an interest in statistical practice, or does it exist only for accredited Disclaimer limit calculation concerning the 24 out of 25 workshop attendants who considered statisticians? The views of contributors to this their workshop lunch excellent. Of course Sincerely yours, Newsletter should not be attributed to the calculation itself could have simply Marks R. Nester the Statistical Society of Australia, Inc. Subscriptions Editorial The Newsletter of the Statistical Society of Australia is supplied free to all The pre-Christmas conference season is behind us now, and reports from several members of the society. Any others wishing to subscribe to the Newsletter recent conferences are in this Newsletter. Letters to the Editor are continuing to arrive at may do so at an annual cost of A$25.00 a steady rate, and more appear in this issue. If you would like to express a view about for an issue of four numbers. a matter of interest to Society members, please write to the Editors at any time. We will typically seek a response from relevant members of the Society, and endeavour to print Advertising letters and responses in the same issue. Central Council and the Canberra Branch have also been busy in the preceding Advertising will be carried in the year, producing posters for displaying and celebrating statistics in Australia. An article Newsletter on any matters which the about the posters appears in this issue of the Newsletter. The Editors hope that you will Editors feel are of interest to the members consider purchasing a set (or several sets!) of posters for your institution’s noticeboards. of the Society. For details of advertising There was a limited print run in 2003, and plans for a bigger and brighter project in 2004. rates, etc. contact the Editors at the above If you have a view to express to the project managers, please contact them – they’d love addresses. to hear from you! Printer National Capital Printing Competition 22 Pirie Street, Fyshwick ACT 2609

The Editors are still seeking entries in the competition announced in the November newsletter. DEADLINE FOR The object is to come up with an adjective that is the most appropriate for a statistician NEXT ISSUE: (or biometrician or whatever). We’re looking for something like a “nonplussed 10 APRIL 2004 mathematician” or a “mean statistician”. The best entries will appear in the next issue of the Newsletter, so send your contributions to the Editors by 10 April 2004.

6 SSAI Newsletter – February 2004 The next Australian Statistical Conference is in Cairns 11–16 July 2004 in parallel with the next International Biometrics Conference IBC2004

Peter Diggle, Patrick Heagarty, Andrew Harvey, Noel Cressie, Yoav Benjamini, Peter Muller, Peter Thompson, Eleanor Feingold, Robert Kohn and Matthew Stephens will be there to talk to you.

Will you be there to hear them?

Register now to save money! Visit www.ozaccom.com.au/cairns2004 to find details of the scientific program, workshops, accommodation, social program and tours, and how you can register.

Early bird registration closes 1 March 2004.

SSAI Newsletter – February 2004 7 Feature: A Career in Statistics

Betty Laby shares with us some her estate, part of which was left moments from a career that spanned to the University of Melbourne to over 30 years working with statistics. commemorate Maurice’s considerable My career in statistics began in 1950, contribution to statistics in Australia. two years after Maurice Belz formed In 1986, my sister Jean and I went to the first autonomous Department of live in our old family home in Mount Statistics in Australia at the University Eliza, about 50 kilometres south of of Melbourne. He introduced a service Melbourne. However, for about 15 within the department to assist other years, we maintained our flat in Kew scientists at the University with the to enable us to attend functions in the analysis of their experiments. My job city with our Melbourne friends. Most was to analyse data for the Statistics years, this included the Belz lecture staff, as well as staff and postgraduate and dinner of the Victorian branch of students from other departments in the SSAI. the University. Betty Laby at home in Mt Eliza. At Mount Eliza, we spend as much At first, I and two research assistants time as possible trying to look after used electro-mechanical calculators, Department of Statistics, and our large garden. We no longer travel but gradually we progressed to continued to assist with selected overseas, but until recently we took electronic machines. At the same projects for the next 10 years, our holidays interstate. Our last trip time, I acquired some knowledge of including the Alcoa study and the was to Mildura in 1998, when Jean statistical methods. preparation of a paper for publication helped a balloon launching group One of our first outside jobs was based on it. I spent one day a week from Laramie, Wyoming, USA, to the analysis of measurements of on this task, and on keeping in touch carry out some atmospheric research birds and seals in several locations with the Consulting Centre. The Alcoa with the CSIRO. in the Antarctic. This topic was paper “A longitudinal study of lung I keep in contact with a few of my among the many disciplines covered function in non-asthmatic workers in old statistical friends. Warren Ewens in our work, spanning Architecture an aluminium smelter” was finally makes time to come and see us on to Zoology. Over the years, our published in 1997. In my last years his visits from Philadelphia. We are consulting service was called upon as an associate, I was an observer at pleased to see Herbert David on his to assist with many research projects; some of the meetings which preceded rather less frequent trips to Australia. these resulted in several postgraduate the merger of the Mathematics and Daryl Daley has family in Melbourne, and staff research publications. My Statistics Departments, and was sorry and so comes down from Canberra help in these was acknowledged when the amalgamation took place. quite often. Evan and Judy Williams with gratitude, or in some cases joint It is nearly 18 years since I retired, live in Geelong, on the other side of authorship. and my only remaining association Port Phillip Bay, but we keep in contact In 1984, a Statistical Consulting with my old colleagues is an annual by telephone. Joe Gani and his family Centre was formally opened at the Christmas lunch with the Alcoa have been our good friends for many University of Melbourne; it continues group. Soon after my retirement, years, and since Joe’s retirement in to supply a service “to promote Jane Matthews became Secretary 1994, we have been able to see him on the practice of appropriate and of the Australasian region of the his regular visits to Melbourne, and informative statistical methods”. I International Biometric Society, and I our less frequent visits to Canberra. In became a member of its staff, and agreed to act as its Treasurer for two 1948 and 1949, we were both working continued to help with the analysis of years. in the Mathematics Department of the various studies referred to the Centre. I was kept busy writing the University of Melbourne. I have boxes One of these, for Alcoa of Australia, Australian Dictionary of Biography of postcards and letters from the Gani was concerned with a longitudinal entry for Maurice Belz, and later family from all corners of the world, study of the lung function of workers looking after Marjorie Belz’s estate, spanning a period of 55 years. in an aluminium smelter. which took nearly three years to Apart from gardening, I have tried When I retired in 1985, I was settle. Marjorie died in 1995 at the my hand at various handicrafts, appointed an associate in the age of 97, and I was co-executor of including patchwork to use up material left over from our dressmaking days. Various relatives Membership Renewals seek help in compiling our family tree, which is quite illuminating. Much less entertaining is the endless Renewal notices were mailed to all members in December 2003. paperwork associated with taxes and Thank you all the members who have already renewed. Please note that if your account keeping: I would enjoy my membership renewal is outstanding at 31 March 2004 you will not receive any retirement more if the various levels more SSAI publications until payment is received. of government did not make life so If you did not receive your renewal notice please contact Jane Waslin at the SSAI difficult. office ([email protected]). Betty Laby Mt Eliza, Victoria 3930

8 SSAI Newsletter – February 2004 Feature: A Celebration of Statistics in Australia

The Canberra Branch, in conjunction contributions that could be displayed on with Central Council and the University a poster, we see the current series as a of Canberra, is delighted to announce pilot project. Next year we would like to that it has completed a pilot project to continue the project. We would now like produce a series of posters for display in to consult with other branches to reach universities, in workplaces, indeed in any agreement on the intended audience, location frequented by people with an poster aims, poster series topics, interest in statistics. individual poster topics, material to be The main aim of the pilot project was included on each poster, and methods of to celebrate the history of statistics in marketing the posters. Possible aims for Australia. The idea arose from a scheme future poster series could include operating at the University of Canberra. 1. encouraging students to study In the first semester of their third year, statistics and become statisticians; Graphic Design students at the university 2. promoting statistics as a profession; design posters for external clients as a part of their assessment. There is a charge 3. raising awareness of important for printing, and a minimal design fee. scientific advances that have been made by applying/developing In early 2003, we realised that the statistical techniques; and SSA could make use of this opportunity. However there was insufficient time to 4. showing the range of the disciplines gain widespread support for the project in which statisticians are employed. amongst the statistical community. A If you have a comment about the current pair of Graphic Design students became set of posters, or would like to discuss involved immediately in designing ideas for future posters, please use the four posters on important Australian discussion board on the SSAI website. Go statisticians. They eventually handed to http://www.statsoc.org.au/~ssacanb the project on to a third student who and follow the link to the discussion completed the work in November. board. The poster series is a celebration The limited print run of posters will of statistics in Australia, focusing on be sold by the SSAI Office at a cost of four important Australian statisticians $30 for the series of four posters. These – Cornish, Knibbs, Lancaster and Pitman. funds will be used to refund the Canberra These men were chosen to represent a Branch and the University of Canberra range of the Australian states and a range for their part of the funding, and support of areas of application/theory, as well as a broader project in 2004. We hope that for the importance of their work. Other every university department in Australia statisticians who were considered for will purchase a set, along with a wide the posters were Belz, Moran, Hannan, variety of workplaces where statistical Wilkinson, Foreman and Watson. methods are in frequent use. With such a variety of people and Ann Cowling and Alice Richardson

Order form – Statistical Society Posters Please supply: 1 set of posters at $30. ❑ 2 sets of posters for $50. ❑ (insert number) _____ additional sets of posters, $20 each ❑

Total cost $ ______A cheque is enclosed/please charge my credit card ❑ Visa ❑ Mastercard ❑ Bankcard Number ______

Expiry date: _____/_____ Name on card ______

Signature of cardholder: ______

Deliver to: ______

______

______(State) ______(Postcode) ______Please photocopy this form and send it to Poster Orders, Statistical Society of Australia Inc, PO Box 4085, Ainslie ACT 2602

SSAI Newsletter – February 2004 9 Springer for Statistics

L. Wasserman W. Hörmann, J. Leydold, G. Derflinger Y. A. Kutoyants All of Statistics Automatic Nonuniform Statistical Inference A Concise Course in Statistical Inference Random Variate for Ergodic Diffusion Written for people who want to learn Generation Processes probability and statistics quickly. The book provides a concise introduction The recent concept of universal random Encompasses a wealth of results from to a larger number of topics that are variate generation can only be found dis- over ten years of mathematical literature. usually included in a graduate-level persed in the literature. Being unique in It provides a comprehensive overview of mathematical statistics class. its overall organization, the book covers existing techniques, and presents - for It brings together many of the main not only the mathematical and statistical the first time in book form - many new ideas in modern statistics in one place. theory but also deals with the imple- techniques and approaches. The book is suitable for students and mentation of such methods. A state-of-the-art reference that will researchers in statistics, computer sci- All algorithms introduced in the book are prove invaluable to researchers, and ence, data mining and machine learning. designed for practical use in simulation graduate and postgraduate students, and have been coded and made avail- in areas such as financial mathematics, 2004. XIX, 442 p. (Springer Texts in Statistics) able by the authors. Examples of possi- economics, physics, mechanics and the Hardcover € 84,95; sFr 144,00; £ 65,50 ISBN 0-387-40272-1 ble applications of the presented algo- biomedical sciences. rithms are presented at the end of the 2004. 490 p. (Springer Series in Statistics) Hardcover book. € 84,95; sFr 144,00; £ 65,50 D. S. Silvestrov ISBN 1-85233-759-1 2004. X, 442 p. (Statistics and Computing) Hardcover Limit Theorems for € 74,95; sFr 127,00; £ 57,50 ISBN 3-540-40652-2 Randomly Stopped A. A. von Davier, P. W. Holland, D. T. Thayer Stochastic Processes W. Härdle, L. Simar The Kernel Method of The first book to present a state-of-the- Applied Multivariate art overview of both theory and applica- Test Equating tions. It covers and demystifies the vast, Statistical Analysis The first presentation of the kernel and technically demanding, Russian liter- A state of the art presentation of the method in book form. Kernel Equating ature in detail. tools and concepts of multivariate data (KE) is a powerful, modern and unified 2004. XIV, 398 p. (Probability and its Applications) analysis with a strong focus on applica- approach to test equating. It brings Hardcover € 109,95; sFr 186,00; £ 60,00 tions. together several methods into an organ- ISBN 1-85233-777-X The text presents a wide range of exam- ized whole rather than treating them as ples and 228 exercises. disparate problems. 2003. IV, 486 p. Softcover € 69,95; sFr 123,50; £ 54,00 2004. XXII, 229 p. 63 illus. (Statistics for Social Science ISBN 3-540-03079-4 and Public Policy) Hardcover € 74,95; sFr 127,00; £ 57,50 Please order from ISBN 0-387-01985-5 Springer · Customer Service Haberstr. 7 · 69126 Heidelberg, Germany Tel.: +49 (0) 6221 - 345 - 0 springeronline.com Fax: +49 (0) 6221 - 345 - 4229 e-mail: [email protected] or through your bookseller All Euro and GBP prices are net-prices subject to local VAT, e.g. in Germany 7% VAT for books and 16% for electronic products. ��� Prices and other details are subject to change without notice. d&p · 010651x

10 SSAI Newsletter – February 2004 Section News

My name is Peter Howley and for researchers across the world describing through retrospective chart review, those of you who didn’t know I was their working environments and and survey based studies. Traditional appointed as the SSAI’s Section Chair experiences as well as perspectives on statistical training led us to believe that of Statistics in the Medical Sciences in the past, present and future research and good statistical practice starts with an mid-2003. I’d like to extend my wishes applications of statistics in the health optimal design, which results in a well- for a positive and fruitful New Year and and medical sciences. The first of many balanced sample that suits exactly the take the opportunity to begin 2004 with forthcoming international contributions desired analysis. However, in health a brief introduction of myself. is from Canada and was written by services research, such an ideal and clean From 1995-98, I worked full-time colleagues employed at the Queen’s dataset is often impossible or impractical as a statistician in the Health Services Cancer Research Institute, Ontario. to obtain. In some other cases, what Research Group (HSRG), Newcastle we are interested in is, in fact, what ‘A brief journey around the whilst lecturing part-time at the happened in the population at large. In globe’ University of Newcastle. The HSRG is these cases, we work with population a multi-disciplinary team that provides Statistics in the Division of Cancer Care and data, in which the unbalanced sample consulting, research and training to Epidemiology at Queen’s Cancer Research sizes across study units, unaccounted support data-based decision making in Institute – A decade in health services confounding and imputed data due to a health care. Major statistical analyses research lack of information are inherited through and reports included the Quality in The Radiation Oncology Research Unit the design. Australian Health Care Study which (RORU) began as a multidisciplinary Along with these challenges, we also assessed adverse events and the health services research group witnessed with excitement the growth of potential for quality improvement in the established in the Department of statistics in health services research, for healthcare system; the Cancer Action in Oncology at Queen’s University in 1992, example, the introduction of hierarchical Rural Towns project which studied the with a mandate to study all aspects of modeling to the field. The new effects of community cancer-prevention the delivery of radiation treatment. In methodologies have provided solutions programmes on smoking behaviours by 1997, the group was designated as a to some above-mentioned issues. At sampling adolescents and adults from Provincial Research Unit within Cancer the same time, they also posted new rural towns in NSW; and an Ethnic Care Ontario, a provincial government challenges and opportunities. We find Study which analysed the association cancer agency of Ontario, Canada. ourselves constantly facing the questions between ethnic background, admission In 2001, the RORU joined with several such as: when to use the new modeling rates and length of stay in hospital. new researchers in cancer care and methods, how to calculate sample size, I have lectured full-time since epidemiology to create the Division of and how to assess model adequacy. 1999 and I am submitting my Ph.D. Cancer Care and Epidemiology (DCCE) The second challenge lays in the (Statistics) thesis in early 2004 entitled of the Queens Cancer Research Institute balance between sound statistical “Analysing and reporting clinical (QCRI). The interests of the Division methods and the interpretability of indicators using hierarchical models”. broadened to include: environmental the results to health care providers The thesis involves research into the and genetic aspects of cancer etiology; and health officials. At the DCCE, use of Bayesian hierarchical models for population-based needs assessment; the statistical development is not for the the reporting of performance indicator management and outcome of cancer at sake of statistics, but for the sake of data for the Australian Council on the population level; access to cancer cancer research. Although as a branch Healthcare Standards and Adverse treatment services; clinical decision- of mathematical science, statistics has its Event and Patient Safety studies. My making; the evaluation of treatment own beauty, it is a useful and powerful broader interests lie with research and guidelines; and the development tool that millions of researchers in applications of statistics in the fields of of public policy related to cancer the applied fields rely on to draw health, medicine, finance, management, prevention and treatment. meaningful inference and conclusions. industry, business and education and The DCCE, headed by Dr. Mackillop, We as statisticians need to select or build quality improvement activity in those has thirteen principal investigators with tools that are reliable, transparent, and areas. I’m currently a chief investigator expertise in radiation oncology, surgical best suited to the purpose. In order to of projects looking into the robustness and medical oncology, palliative care, achieve this goal, a good biostatistician of newly developed reporting and epidemiology, psychology, health also needs to be an expert in their analyses techniques for clinical indicator economy, and medical physics. There are applied field. Regardless how advanced data, the use of employee attitude currently four biostatisticians/statistical statistical methodologies become, this surveys in local business and industry analysts who provide statistical will remain a challenge. and improving the reporting of the expertise and statistical computing skills by Jina Zhang-Salomons, Patti Groome University’s performance indicators. to support the research program. and William J. Mackillop. I’ve presented two-day seminars on Drawing from a decade-long continuous improvement to businesses, experience, we describe the life of a I hope that you enjoyed this article including an aged care organisation. statistician in health services research and trust that you look forward to future I would also like to take this as challenging and exciting. The first instalments. opportunity to commence an article challenge is associated with study which I will continue through ensuing design. The studies conducted in the All the best and remember, without editions of the newsletter. The article, DCCE employ a variety of study designs data you are just another person with entitled ‘A brief journey around the such as population-based studies, an opinion. globe’, will feature profiles from sample based case-control studies Peter Howley

SSAI Newsletter – February 2004 11 Member Advantage

WHAT YOUR HEALTH FUND CAN DO FOR YOU Member Advantage Health, in partnership with IOR Health Benefits, offers many special benefits for SSAI members that can’t be found at other health funds.

❖ Up to 5% discount off the approved rates for members ❖ Waiver of the normal two month waiting period1 ❖ All excess waived for day surgery with no overnight stay2 ❖ All excess waived for accidents2 ❖ Save on your premium with the option of daily excess3 ($50 per night) ❖ Family cover, which covers dependant children up to 22 years on age. ❖ We offer security and peace-of-mind and a simple to understand product range.

Top Plus Hospital Advanced Saving hospital cover, excess $50 x 8 hospital cover, excess $50 x 8 State Weekly Rates* Weekly Rates* Family in NSW / ACT $23.65 $11.95 Single in NSW / ACT $11.80 $5.95 Family in QLD $22.85 $13.65 Single in QLD $11.40 $6.80 Family in VIC $22.80 $15.65 Single in VIC $11.40 $7.80

For a free quote call Member Advantage Services on 1300 853 352 or log onto http://www.member-advantage.com/ssai

1. Does not cover pre-existing ailments. A 12 month waiting period applies to these. Waiting periods of 2 months are waived. Some conditions apply. 2. IOR will waive the excess for all same day procedures or for hospital treatments which are the result of an accident. The excess is waived if treatment is within 12 months of the accident. 3. Excess is per person, per night, per calendar year. * The hospital rates quoted do not include Lifetime Cover loading. It does include Federal Government 30% rebate and SSAI member discount of 5%.

Car Hire – Special Rates for SSAI Members Member Advantage and Hertz offer you special member rates every time you rent a Hertz car in Australia and New Zealand. Your car hire discounts also extend to the Hertz Prestige Collection in Australia. Indulge yourself this summer as you unwind behind the wheel of the sleek new BMW Z4 or Volvo C70 convertible.

As a member of the SSAI, you’re also invited to join the Hertz #1 Club Gold at no charge. With your details stored electronically, you’ll no longer have to face paperwork or queues to collect your Hertz vehicle. Plus your special SSAI rates will automatically be applied to every booking you make.

Call Hertz today in Australia – 13 30 39, in New Zealand – 0800 654 321 and quote The SSAI Member Advantage Customer Discount Program (CDP) Number – 1250381 or visit http://www.member-advantage.com/ssai

12 SSAI Newsletter – February 2004 The 2003 Ken Foreman Lecture by Ken Brewer

At the monthly meeting of the Canberra Branch of the SSAI on 14 October 2003, Dr Ken Brewer of the ANU presented the 2003 Ken Foreman Lecture. Ken Brewer (KB) began by expressing his deep gratitude to Ken Foreman (KF), for whom he worked at the ABS from 1954 to 1974. This was an immensely stimulating time throughout, in which KB learnt a great deal. KB has many fond memories of KF, including his saying: “I wouldn’t even scratch myself without carrying out a pilot test first.” In a recent book (Brewer, 2002), KB has paid tribute to KF by modelling one of that book’s principal characters Robert Clark and Ken Brewer on him. One of KB’s early projects under KF was developed further in JRSSB (Foreman support and ready collaboration of KF. the sample redesign of the ABS’s three and Brewer, 1971), and two decades He concluded by leaving the audience Business Surveys: Capital Expenditure, later KF published a monumental book with a thought which he had never Stocks, and Labour Turnover. In one on survey methods in which the topic heard from KF’s lips but which was is treated at length (Foreman, 1991). of the Capital Expenditure Survey’s exemplified by almost everything he industries he came across a group KF’s attitude was in stark contrast to did: “Take your time to think about of several extreme observations. The that of Morris Hansen, who abruptly what you are trying to achieve. The practice in those days was simply to ended a conversation with KB in 1966 at more you know and understand about ignore an extreme observation, unless the US Bureau of Census when he realised it, the better you will be able to tackle it it was extraordinarily extreme, in which that the topic was population models. in the long run.” Following the lecture, case the standard procedure was to delete But KB was not alone in his treatment by KB’s discussant Dr Robert Clark of the it from both the sample and the sample Hansen. Apparently, Morris Hansen and ABS provided an illuminating overview frame and add it on at the end with a Richard Royall almost came to blows of model-based inference from his own sample weight of unity. KB did not feel several times in the 1970s, because of perspective, including examples of comfortable with either of these options, their violent disagreements over the where this approach has had or is likely which he saw as potentially causing a usefulness of model-based inference. to have the greatest impact on survey significant bias. So he investigated the Interestingly, the idea of a population sampling. model goes as far back as 1938 when it matter and found that the extremes in Borek Puza was used for analysing soil samples at question all corresponded to financial 16 October 2003 institutions. This led him to create a the CSIRO in Canberra (Fairfield Smith, separate industry under this heading 1938). Nowadays, population models References can be found almost everywhere in the which was then completely enumerated. Brewer, K.R.W. (1963). Ratio estimation Two similar groups of extremes were sampling literature, for example in the and finite populations: Some results found in the Stocks Survey. Experiences classic book by Särndal, Swensson and deducible from the assumption of such as this taught KB the importance Wretman (1992). an underlying stochastic process. of understanding the nature of data Throughout his career KB has Australian Journal of Statistics, 5, 93- collections and also the way in which devoted much time and thought to 105. potential respondents organise their the relationship between design-based business. and model-based inferences. He sees Brewer, K. (2002). Combined Survey Sampling Inference: Weighing Basu’s In the same project, KB came across that each has its merits. In particular, Elephants a variance anomaly whilst attempting design-based inference becomes more . London: Arnold. to improve on the common practice and more realistic as both the sample Fairfield Smith, H. (1938). An empirical of treating a completely enumerated size and population size increase. On the law describing heterogeneity in the stratum with nonresponse as though other hand, model-based inference may yields of agricultural crops. Journal of it were a sampled stratum, and the be more useful when the population Agricultural Science, 28, 1-23. and sample are small, as often occurs respondents as though they had been Foreman, E.K. (1991). Survey Sampling for small areas, even where the overall randomly selected. In an attempt to solve Principles. New York: Dekker. the anomaly, KB devised the notion of population and sample are large. Foreman, E.K., and Brewer, K.R.W. (1971). conditional variance under a prediction Moreover, there are advantages in using The efficient use of supplementary model, and a few years later published both approaches together, and KB has information in standard sampling a paper featuring that idea in the then over the years devised several hybrid fledgling Australian Journal of Statistics estimators with very attractive properties. procedures. Journal of the Royal (Brewer, 1963). In the mind-set of the KB emphasized that this drawing Statistical Society, Series B, 33, 391-400. 1960s, this idea was very new, and it together of the two inferences, together Särndal, C.-E., Swensson, B., and was a measure of KF’s open-mindedness with various spin-offs, would not Wretman, J. (1992). Model Assisted that he encouraged KB to submit have seen the light of day had it not Survey Sampling. New York: Springer- for publication. Later, his idea was been for the consistently sympathetic Verlag.

SSAI Newsletter – February 2004 13 55th ISI Session Sydney, 5-12 April 2005

Preparations for the 55th Session The Scientific Program for the 55th environment and genomics will in Sydney 5-12 April 2005 are well Session of the ISI includes Invited be the special themes for the 55th underway. The 2005 ISI Session will Paper Meetings, Contributed Paper Session. provide an opportunity for delegates Meetings, Poster Sessions, Short The Invited Paper Meetings have Courses and Tutorials. The Session to exchange new ideas, develop new been developed by the Program will also provide theme days for contacts, and discuss current trends Committees of the ISI and its Sections. those interested in particular topics. and developments in Statistics and its The following is a partial and tentative Statistics relating to finance, the relevance to real world problems. list of invited paper meetings (IPM).

Partial and tentative list of invited paper meetings (IPM):

1. President’s IPM 23. Functional data analysis in quantitative finance 2. Best papers from developing countries 24. Optimisation algorithms for experimental designs 3. Random matrices and methods for high dimensional 25. Statistical learning from data data 26. Computational tools for microarray analysis 4. Recent developments in financial econometrics 27. Statistical environments in the network age 5. Nonparametric methods for functional data 28. Computational advances based on the EM algorithm 6. Nonparametric methods for structural econometric 29. Pattern recognition in high dimensions models 30. Introduction to technical aspects of DNA microarray 7. Errors in measurement: recent advances experiments 8. Stochastic networks 31. The PLS (Partial Least Squares) approach in data 9. Inverse problems and functional estimation analysis 10. Recent developments in joint modelling of longitudinal 32. The use of simulation in statistics education and survival models 33. Reasoning about variation 11. Sampling methods for animal populations 34. Teaching statistics on-line 12. Local parametric modelling for curve estimation 35. Statistical literacy 13. Estimation of the support and efficiency frontiers 36. Quality assurance in statistics education 14. Internet tomography 37. Ethical standards in statistics education 15. IAOS forum 38. Using history of statistics to enhance the teaching of 16. Impact of the “international indicators of development” statistics movement on national statistical programme priorities 39. Developments in the analysis of longitudinal survey 17. Statistical measurement issues requiring collaboration data among NSOs 40. Use of model diagnostics in survey sampling 18. Statistics on international migration 41. Calibration in practise 19. Response burden and response rates 42. Quality measurement and reporting for surveys 20. The role of official statistics in innovation, knowledge 43. Resampling methods for variance estimation in management and development of the new economy complex surveys 21. The regional and urban dimension of official statistics: 44. Experiences in data collection with internet surveys small area statistics and data of particular relevance to regional and urban planning (SCORUS item) 45. Inferential potentials of non-probability samples 22. Standards for regional and urban indicators (SCORUS 46. Confidentiality protection in national statistical offices item) 47. Surveys of small and medium sized enterprises

14 SSAI Newsletter – February 2004 Members of the 2005 ISI Organising Committee after their recent meeting in Sydney. L-R: Steve Dangaard, John Struik, Siu-Ming Tam, Anna Poskitt, Roslyn McLeod, Jonathan Palmer, Dennis Trewin, Graeme Hope, Felicity Kent, Nick Fisher, Sara Foda and Geoff Lee.

Provisional Program

Morning Early Afternoon Late Afternoon Evening Registration Registration Monday 4 April Short Courses Short Courses Short Courses Registration Registration Opening Ceremony * Welcome Reception * Tuesday 5 April Short Courses Short Courses Wednesday 6 April Scientific Meetings Scientific Meetings Scientific Meetings Optional Social Event Thursday 7 April Scientific Meetings Scientific Meetings Scientific Meetings Australiana Night * Friday 8 April Scientific Meetings Scientific Meetings Scientific Meetings Optional Social Event Saturday 9 April Scientific Meetings Excursions Excursions Sunday 10 April Excursions Excursions Excursions Monday 11 April Scientific Meetings Scientific Meetings Scientific Meetings Farewell Party Tuesday 12 April Scientific Meetings Scientific Meetings Scientific Meetings

* Functions included in Registration Fee

More details on the scientific program will be listed on the 2005 ISI website at www.tourhosts.com.au/isi2005 as they are developed.

Alternatively, register your interest by contacting: ISI 2005 Conference Managers Register your interest now! GPO Box 128 Register your interest in the 2005 ISI Session by Sydney NSW 2001 completing the expression of interest form on the 2005 ISI Australia Session website www.tourhosts.com.au/isi2005 Phone: 02 9248 0800 Fax: 02 9248 0894 Email: [email protected]

SSAI Newsletter – February 2004 15 NATIONAL SYMPOSIUM ON PROBABILITY AND ITS APPLICATIONS

ANU Mathematical Sciences Institute 22-23 April 2004

A National Symposium on Probability and its Applications, in honour of Chris Heyde’s 65th birthday, is to be held at the ANU Mathematical Sciences Institute on 22-23 April 2004. Several speakers from Australia and overseas will be presenting papers. If you would like to attend, please let Joe Gani know at [email protected] We hope to hold a banquet on the evening of 22 April in honour of Chris Heyde, at which he will be presented with the Festschrift which Eugene Seneta and I have been preparing for him. The venue for this event has yet to be decided, and may depend on the number of people who wish to take part in it. Please let me know if you would like to attend the banquet; the cost is likely to be around A$ 30-35 per person. If you have any queries about the arrangements for the Symposium, I shall be glad to respond to them. You will find details posted at http://www.maths.anu.edu.au/events/sy2004

Statistician/Biometrician (consultant) Market Research company specialising in Sensory Research and product testing requires a consultant statistician/biometrician. The person we seek must be proficient in multivariate analysis, be able to demonstrate skills in consumer mapping and segmentation techniques. We envisage this to be an ad hoc position depending on the company workload. Initial enquires can be made to Dr.Cathy Gair on (02) 9894 1188, or email summarised resume to [email protected]

16 SSAI Newsletter – February 2004 BioinfoSummer

“Mathematical, Statistical and Biology; Sequences & Data; Evolutionary Simon Easteal (ANU), Warren Kaplan Computational Challenges in Models & Genetics; Protein Structure & (Garvan Institute, Sydney), Mark Ragan Bioinformatics” was the theme of Function and Microarrays & Experimental (IMB, Brisbane), Gordon Smyth (WEHI, the extremely successful Australian Design, the last theme being held jointly Melbourne), Sue Wilson (ANU). A poster Mathematical Sciences Institute (AMSI) with the IBS Australasian Region’s session was held on Tuesday evening, Summer Symposium in Bioinformatics, biennial meeting. Each day started and student scholarships were awarded hosted by the Australian National with Educational Lectures followed by to 28 successful applicants. The ceremony a mix of Keynote and Specialist talks. University’s Centre for Bioinformation was followed by a well-attended public Most days had parallel sessions in the Science, December 1-5. Sponsors also lecture by Jenny Marshall-Graves (ANU) included Cray, Ceanet, the Australian afternoon, allowing those students who on “Unravelling the Kangaroo Genome”. Partnership for Advanced Computing had registered for the Graduate Course NIB prizes were awarded to Ann Kwan and ANU’s National Institute for Award in Bioinformatics to do the (best poster), Penny Bennett and Natalie Bioscience (NIB). There were ~150 Educational Computer Lab session while Thorne (best student talk, shared) and registrants. The days were organised into others attended Specialist talks. Jim Stankovich and Antonio Reverter Themes, allowing attendees who were The keynote speakers were John unable to attend the whole week to select Mattick (Institute for Molecular (best talk by a researcher within ten years those areas of particular interest. The Bioscience, Brisbane), Terry Speed (Walter of receipt of PhD, shared). Themes were: Introduction to Molecular & Eliza Hall Institute, Melbourne), Susan Wilson

SSAI Newsletter – February 2004 17 Branch Reports Continued

QUEENSLAND

Members of the Queensland Branch organised a conference to encourage less experienced statisticians and students to present their work in an informal venue. Another aim of the conference was to give members outside of Brisbane an opportunity to contribute to the Branch, which is difficult with the traditional monthly meetings that are held in Brisbane. The Toowoomba campus of the University of Southern Queensland supported this friendly ambience. Local organiser Peter Dunn put in a sterling effort to organise a web site, conference administration, the conference dinner, sponsors, program, conference satchels, even Attendees warm up during morning tea morning and afternoon teas. Peter did get some moral support from and Rodney Ellis (UQ) who won frames, the “Collection District”, other committee members and more a $50 book voucher from Pearson which represents approximately practical help from USQ students Education. 220-250 dwellings. The new topics Melissa Buckley, Helen Nkansah, The informalities continued with for the 2001 Census were regarding Daniel Burrell, Anita Fredericks and the conference dinner at Gaby’s ancestry, computer use at home, Andrew Powierski. Some attendees restaurant with a mystery after- internet use and schooling. Census even suggested Peter arranged dinner speaker, and an impromptu data is divided into two sets of the dramatic fog that engulfed “Guess-the-Mathematically- variables, those pertaining to people Toowoomba for the entire first day. Important-Numbers-Arranged-In- and those pertaining to dwellings, Super effort Peter. A-4-By-4-Square” game adjudicated with connections relating these two Prior to the conference, John by Bob Murison and won by Rodney as family relationships. Maindonald presented an R workshop Beard. Not sure about the relevance Greg explained the differences on Wednesday, 1st October. Over 20 of the 4 by 4 design but it seemed like between the Australian Standard participants worked through a well- a good idea at the time. Geographical Classification (ASGC) structured practical session run by Sponsors of the conference were and the Census Geographic Areas John. In addition to John’s input, Duxbury, SPSS and CSIRO. (CGA) which are approximated by three other speakers, Ross Darnell, postcodes. The former is a hierarchy It is hoped that the conference Peter Baker and Peter Dunn gave incorporating the Access/Remoteness can be run every two years as not presentations on specific R packages, Index of Australia (ARIA) increasingly to clash with the SSAI national lme, Bioconductor, and Sweave. used by health and social researchers. conference. The conference was The conference featured the Journey of work areas were an successful because of the effort of guest speakers Kerrie Mengersen interesting topic based on connections many people and their willingness (“Adjusted likelihoods for meta- between these CGAs. to give some of their time for their analysis: Effects of environmental Features of the 2001 Census showed tobacco smoke”), John Maindonald colleagues. The conference returned a small profit to the Branch. I would an aging population, more than one (“Predictive validation issues in data quarter of Australians born overseas mining”), Dave Butler (“SAMM: An S- like to thank the Branch Council for and decreasing households. Tasmania Plus module for mixed models using supporting the event and particularly showed a drop in population from REML”) and Bob Murison (“Where the promotional effort of Bronwyn the 1996 Census. Greg defined ASREML is best”) and 16 contributed Harch. Again I would like to thank the parameters used in functions talks, six of which were from students. Peter Dunn for a superb effort in to estimate the current resident The range of topics was diverse as organising a great conference. population and flagged the imminent were the applications, including talks November meeting 20 million mark. on agriculture, econometrics, health, environmental and industrial quality Greg Lawrence, Client Manager Of increasing importance is the control. A full program can be found with the Australian Bureau of aging population trend and Greg at the Conference web-site at http: Statistics, presented an entertaining demonstrated how this varies //www.sci.usq.edu.au/staff/dunn/ sample of the 2001 Australian census. significantly with locale. qstatconf/index.html. The winners of Greg discussed the primary reason Approximately 23% of households the student talks were: Ben Stewart- for doing a census, a ‘head count” at were single person households. Also Koster (Griffith University) who won the National, State, local government 66% dwellings were fully owned or $250 from the Queensland Branch areas and the smallest of sampling being purchased compared to 60%

18 SSAI Newsletter – February 2004 Branch Reports Continued

in 1991, while the current monthly housing loan was $870 and the median monthly rental was $616. Internet usage varied from 40-50% in the capital cities and much lower 10-20% usage in the rural areas. Greg then took the opportunity to market ABS products and services. Discussions about these continued at a local Thai restaurant.

December Meeting The final Branch meeting for 2003 was a presentation by SSAI President Neville Bartlett on issues surrounding the Australian and New Zealand Journal of Statistics going fully electronic. A discussion followed between the SSAI President and members of the Qld Branch. Much of this topic has been discussed elsewhere and the issues raised by members were noted by Neville. The Christmas dinner held at Amphora’s restaurant was well attended. R Darnell John Maindonald delivers his talk Vale: John Douglas Kerr 1942-2003

On 26 November 2003, John Kerr for visually estimating pasture yield sadly passed away after collapsing and composition. The computer from a heart attack while riding his program which John wrote was pushbike. Members of the Queensland widely used throughout CSIRO Branch of SSAI will recall John arriving and the Queensland Department of at meetings helmet in hand. John was Primary Industries by researchers a very active participant in many undertaking grazing and agrostology spheres of life. trials. A lot of John’s statistical analyses After winning an Open Scholarship were carried out using a statistical to the University of Queensland, package he had developed himself John completed his B.Sc. (Hons) in (FORALL) and which preceded many 1963. In 1964, he worked briefly of the commonly available packages with ABS and then as a biometrician of today. with the Queensland Department Always a confident and obliging of Forestry, before joining CSIRO consultant with all his clients, he also in 1965 to consult on statistics with had a generous nature often donating researchers of the Fisheries and Marine Dr Rob Sutherst and his colleagues on to the University of Queensland. Division. However, for most of his research into the effects of ticks and tick Upon retirement from CSIRO, John career in statistics, John worked as a borne diseases on livestock productivity effectively started a new career as a consultant biometrician at the CSIRO and methods of control. Even after his historian and wrote extensively on the Long Pocket Laboratory, Brisbane. retirement from CSIRO in 1997, John sugar industry, Queensland regional During that time he made a very continued to collaborate with Rob, and history, railways, ports and mining. significant contribution to the research together with colleagues in Zimbabwe He undertook detailed research on undertaken there and has over 50 they extended their work in Australia to each topic and published 13 books in scientific papers to his credit. John African conditions. these diverse areas, adding to his early provided valuable statistical advice John’s statistical expertise was not historical papers about statistics in on parasitological and entomological confined to methods for analysing the sugar industry and the statistical research into buffalo fly and cattle insect and parasite data. With John issues facing marine sampling. tick and also the biological control of Hargreaves and his colleagues, he John will be sadly missed and the water weeds. This included a long developed a comprehensive sampling condolences of all members of SSAI go and productive collaboration with and computing procedure (BOTANAL) to his wife, Dr Ruth Kerr.

SSAI Newsletter – February 2004 19 Branch Reports Continued

CANBERRA experiences with time dependent through Robins. To add to the sense count data. One of Dr Stern’s past of occasion, the audience of usual projects involved measuring the suspects was supplemented by a Talk on time dependent ages of rocks using a machine bubbling horde of young and eager counts by William Dunsmuir called SHRIMP which shot ions epidemiologists. into rocks, and then analysing the In the early days of HIV and At the Knibbs lecture 25 resulting counts using the Poisson AIDS, when only one or two drugs November 2003, Professor William distribution. In his rejoinder, of promising but unproven efficacy T.M. Dunsmuir of the University of William mentioned that goodness were available, a few randomised NSW gave a talk titled “Estimation of fit tests under serial dependence controlled trials (RCTs) were and Modelling in Time Series of have not been sufficiently developed undertaken. These demonstrated Dependent Counts”. William did and suggested this topic as worthy that the proposed treatments his PhD in statistics at the ANU in of future research. did slow the progression of the the 1970’s under Ted Hannan and disease, as hoped. Today RCTs are Pat Moran. infeasible: numerous drugs and drug combinations are available, The central theme of William’s VICTORIA and few can be compared in RCTs; talk was that incorrect inferences the medical treatment is customised may result if serial dependence in Estimating the causal effect to each patient so that randomised time series is not properly accounted of HAART (highly active allocation is unethical; and greatly for. One of his motivating examples antiretroviral therapy) in the improved patient prognosis means involved a statistical analysis of Swiss HIV cohort study that modern trials would take a long deaths from heroin overdoses in time and be very expensive. Even the NSW in a report on the medically Epidemiologists rarely attend best-designed modern studies give supervised injecting centre at Kings Victorian Branch meetings because rise to time-dependent confounding Cross. Another example involved a it is so hard for them to resist the between treatment and covariates: study of the effects of factors such attractions of MEG – the Melbourne for example, antiretroviral therapy as airborne pollution, humidity Epidemiology Group, which holds raises the CD4 cell count (a low and school terms on daily asthma regular topical seminars. So the cell count indicates that a patient’s presentations at Campbelltown, September Branch meeting, held immune system is weakened by Liverpool and other hospitals in of course on October 7, took on a HIV), but the level of treatment is Sydney. Yet another example had governed by the same cell count. to do with assessing the impact of Traditionally, such confounding is alcohol policy on violent deaths, handled by statistical tools such as including suicide. One other stratification or regression. illustration involved counts of purchases of alcohol by ‘youthful- Jonathan described an alternative looking persons’ without proof of procedure, based on the marginal age (research workers), and the structural models of James Robins question of whether a training of Harvard University, that enables program had improved the practices causal inferences to be drawn from of staff in this regard. modern cohort studies. Imagine, if you can, a trial with two time William has used the framework points. It has many of the features of generalised state space models of a randomised controlled trial, but to address the problem of time has been modified so that treatment dependence in count data. These and a covariate are confounded. models may broadly be categorised At the first time point a patient as parameter driven or observation is allocated either treatment or driven. Generally, the theory for placebo with probability ½. At the analysing parameter driven models Jonathan Sterne second time point, treatment or is simpler, but observation driven placebo is again allocated randomly: models (also known as transition the probability of treatment remains special standing when Jonathan models) are easier to apply. William at ½ if the CD4 cell count is high, Sterne, of the Department of Social has compared many different but rises to 0.9 if the cell count is approaches to dealing with count Medicine at the University of Bristol, low. The conditional probability of dependence including those of gave the presentation. Jonathan was a high CD4 count at time 2 is, say, Zeger (1988), Chan and Ledolter visiting John Carlin, the irrepressible 0.3 under treatment at time 1, and (1995), Kuk and Cheng (1997) and principal medical statistician at 0.1 under placebo at time 1. The Davis and Rodrigues-Yam. the Royal Children’s Hospital situation can be fully described by William’s discussants were Melbourne, for three months. It a probability tree diagram with 8 Professor Ross Maller and Dr Steven gradually became clear throughout terminating nodes, corresponding Stern of the School of Finance and the evening that Jonathan and John to the 8 groups of patients: Applied Statistics at the ANU. both have a Harvard connection, treatment or placebo at time 1, Both recounted some of their own John through Rubin and Jonathan low or high CD4 count at time 2,

20 SSAI Newsletter – February 2004 Branch Reports Continued

treatment or placebo at time 2. In Talks, such as this one by Vladimirovich Gnedenko, with contrast, there are only 4 treatment Jonathan Sterne, that disseminate whom he has a direct professional regimes: treatment or placebo at and popularise new methods at an association, as our windows into time 1, treatment or placebo at time accessible level are very important. the drama. 2. According to Robins, the only Robins has managed to convince Khinchin (1894-1959) studied substantive difference between this journal editors, colleagues and under the real analysts D.F. trial and an RCT is the number students that his ideas have merit. Egorov and N.N. Luzin at Moscow of patients in each group. (In the It is time for them to be subjected to University. He found himself in RCT the data are classified by the the blowtorch of wider experience learned company. Luzin invited 4 treatment regimes – the 8 groups and critical scrutiny that the general students to weekly intellectual are latent, but they still exist.) It is statistical community can provide. evenings at his home, where they a trivial matter to adjust the group Geoff Laslett engaged in lively discussions of numbers in the non-RCT trial to mathematics and partook of tea those of an ‘equivalent’ RCT with Teacher and student: and nut-cake by the light of a equal numbers of patients in each glimpses into the lives and kerosene lamp. In 1927 Khinchin treatment regime. The adjustment times of A. Y. Khinchin and was appointed Professor at factors are called inverse probability B. V. Gnedenko Moscow University, where he weights. Causal inference can established fundamental theorems now be carried out by adjusting It is a fact not well-enough known in probability and stochastic the outcome variables using these that the Russian mathematicians who processes. He held strong views weights, Robins argues, just as if a founded modern probability theory on mathematical education. He genuine RCT had been conducted. did so at grave risk to their lives. was particularly critical of ‘the Jonathan Sterne applied this idea The moving story of the political widespread tendency for students to a Swiss HIV cohort study with dangers facing these pioneers was to learn only the outer, formal 12000 patients, 93000 follow-up the essence of the 2003 Belz Lecture, content of a topic, without ever visits, 36000 diagnoses, 11000 AIDS sympathetically presented by David understanding its real meaning’. events and 4000 deaths. Of course, Vere-Jones, Emeritus Professor of He recommended his own method this is a much more complicated Statistics at the Victoria University of learning new material, namely study than the didactic example of Wellington, to a packed house to put aside the source document, of the previous paragraph: there at the October Meeting of the and to try to reproduce the ideas, are many more time points and Victorian Branch. Although many and extensions of them, using possible treatment regimes. Hence gifted Russian mathematicians were new notation and different steps the inverse probability weights are involved, David chose Aleksandr from those of the original author. estimated using a logistic model Yakovlevich Khinchin and Boris During the 1960s David helped rather than non-parametrically. Even so, causal inference using Robins’ marginal structural model is unreliable because some group probabilities are very small, leading to huge inverse probability weights. Jonathan Sterne produced, as if by magic, some alternative stabilised weights to use instead of the inverse probabilities. These have been recommended by Robins, but Sterne did not explain their intuitive basis. At this point the troops became restless. No verbal tomatoes were actually thrown, but a couple of audience members were spotted ordering crates. From my follow- up reading of Robins’ papers, it appears that the stabilised weights adjust to an RCT with unequal numbers per treatment regime, the same as those actually realised, so they are not too rocky. Sterne showed us how to fit marginal structural models using the statistical package Stata – it takes only a few lines of code. He compared the results for the Swiss HIV study with those from more A walk on ‘The Wild Coast’. David Vere-Jones at Blairgowrie Back Beach a few days conventional, and biased, analyses. prior to the 2003 Belz Lecture. Photo: Brian Phillips.

SSAI Newsletter – February 2004 21 Branch Reports Continued prepare an English translation of and to develop confidence in their Khinchin’s educational articles. reasoning skills. Despite his previous This work appeared in 1968 under experiences at the hands of officials, the title ‘Essays in Mathematical Gnedenko often quoted from the Education’. canons of communist literature when But ideologues were in power. these writings supported his own They were suspicious and beliefs. For example, he cited Lenin intolerant. Independence of in support of the need to include thought, the life-blood of academics statistics in the school programme. like Khinchin and his colleagues, On one occasion Gnedenko took was seen as a threat. In the late a group of former students to the 1920s and early 1930s the Moscow Lenin Museum outside Moscow, to look at Lenin’s original writings Mathematics Society was accused of on applied mathematics. In 1967, being a centre for anti-communist Gnedenko visited Australia (including propaganda. Khinchin’s teacher, Melbourne) at the invitation of P. Egorov, was a staunch supporter Moran and H. Thomson, an event that of academic freedom and the Russian Orthodox Church. He was several in the audience could recall. denounced, deported and ultimately Both Khinchin and Gnedenko Kathy Ruggiero died after going on a hunger strike. were survivors in an extremely For five years the situation hung treacherous political environment. and effect, that is, the responses in the balance. Luzin, the next They had to walk a tightrope in observed between different scientific leader, was accused of which they preserved their own interventions are indeed due to liaising with European intellectuals integrity while acknowledging the those applied interventions. with the aim of engineering a need to work within the system. It Kathy pointed out that the counter-revolutionary plot. At was truly an Age of Uncertainty, experimental design paradigm is the last moment the charges were in two senses: political survival; needed to invoke cause and effect dropped, but he spent some time in and advances in the theory of inference. She then showed how the a labour camp before reappearing in probability. experimental design paradigm can the 1940s. It was a very successful evening. be practically invoked with cDNA Gnedenko (1912-1995) worked from David’s intimate and sensitive microarray experiments, to provide 1933 to 1937 as a graduate student at portrayal of the Moscow school of efficient cause and effect inference. Moscow University with Khinchin probability in its perilous formative This involves knowledge of possible and A.N. Kolmogorov. In 1938 he years enthralled the large audience major sources of variation. was arrested, held for many months of over 70 people. Nearly 40 went The invoking of the experimental in grim conditions, and subjected to on to an animated and convivial design paradigm leads to major daily and forceful interrogation, with Belz Dinner, held at the Universal differences in the way microarray the aim of getting him to ‘confess’ Café in Lygon Street. experiments need to be designed that Kolmogorov was involved in Geoff Laslett and analysed, compared to the an anti-communist plot. He feared predominant present approaches. that if he gave in, the whole Moscow An old paradigm for a new The paradigm leads to row-column school of probability would be lost. technology: designs for cDNA designs which do not require He insisted that the claims were microarray experiments dye swaps of technical replicates and replaces the within array false, and was ultimately released. Kathy Ruggiero, of CSIRO normalisation with normalisation by Kolmogorov and Khinchin were Mathematical and Information design using an intrablock analysis able to secure his reinstatement in Sciences in Canberra, spoke on of variance. Further, it eliminates academia, but Gnedenko remained designs for microarray experiments the need for control and/or forever debarred from the highest at the November meeting of the housekeeping genes. honours. Gnedenko went on to do Victorian Branch. This work was much original work on sums and initiated in Melbourne while Kathy Kym Butler maxima of sequences of random was on sabbatical from the Institute variables, on reliability and queuing of Information and Mathematical theory. In 1960 and 1961, David Sciences at Massey University, and SOUTH had the privilege of working in I had the privilege of collaborating AUSTRALIA Moscow as a graduate student under with her. Gnedenko’s supervision. The technology of cDNA Practical analysis of spatial Like Khinchin, Gnedenko held microarrays is often used to point patterns influential views on mathematical compare differences in gene education. He believed that the expression between interventions Professor Adrian Baddeley, Professor relationship between teacher and on many (usually thousands) of of Statistics at the University of Western student was paramount – teaching genes simultaneously. In such cases Australia. mathematics means helping young the investigator would often like Adrian kindly delivered the second people to reason for themselves, to make statements inferring cause

22 SSAI Newsletter – February 2004 Branch Reports Continued

Cornish Lecture to the SA Branch on 21 point processes is analytically intractable parametric/non-parametric methods October 2003. The lecture series has been and achievable with computationally (robust statistics) in econometrics such named to commemorate Alf Cornish, a intensive methods. Adrian outlined as generalised method of moments leading figure in the early years of the use and flexibility of the and bootstrapping. Nevertheless, the statistical profession in Adelaide. pseudolikelihood. It is formally identical regression techniques still have a The lectures are held biennially and to the likelihood of GLM (binomial or strong hold in economics. The number presented by eminent statisticians from Poisson log-linear regression) allowing one journal read by most economists, around the world. The first Cornish use in standard software for GLM. The American Economic Review, publishes Lecture was delivered by Professor software R has been useful in fitting such many papers that use standard linear Terry Speed of the Walter and Eliza models and also exhibits flexibility. regression (at the exclusion of other Hall Institute of Medical Research and The speaker and topic attracted a methods). In contrast Econometrica the Department of Statistics, Berkeley, large crowd who were drawn from is filled with leading edge statistical on the subject of ‘Gene Expression’. a wide sector of the mathematical methods. This position is slowly The statistical analysis of spatial point and statistical community. The talk changing. One of the most heavily cited patterns is an important and challenging generated many questions and papers in recent times being Halbert task and has many applications (e.g. in discussion after which a large group White’s 1982 paper, “Instrumental epidemiology, agriculture, ecology and adjourned to a local restaurant to Variable Regression with Independent materials engineering). It is challenging continue the lively chat. Observations” (Econometrica). This because it is methodologically different paper is significant in the development from most other areas of statistics. Margaret Swincer of White (or Huber-White) standard Statistical methods for point pattern errors, which are standard errors Econometrics – where to data are relatively underdeveloped; adjusted for specified assumed-and- from here? the classical methods are very limited estimated correlations of error terms in scope. Dr Tony Hughes of the Department of across observations. Furthermore, the observation points Econonomics, University of Adelaide. Economists use a combination of non- of point patterns are not actual data The original concept of econometrics sampling information and econometric points as information is conveyed by was a “synthesis of statistics and methodology. For example if there is a the absence of points. mathematics applied to economics” strong prior belief or economic theory Point patterns can be a binary and traditionally focussed on structural predicts that a particular coefficient (0/1) variable if we have a rectangle analysis and forecasting. is positive say, then this presumed of squares. Models for point patterns An expert econometrician in the 60’s structure is built into the model. include: inhomogeneity over space, and 70’s could run a regression whereas This approach has the advantage inter-point interaction (if none is today this has evolved to understanding of providing a lot of efficiency and present we have a Uniform Poisson a regression. Econometrics is now power, but at the possible expense of process) and covariate effects. thought of as Economic Statistics where robustness. Technical problems include: one- the distinction between econometrics In the 1940’s governments around dimensional time series have no natural and statistics is the role economics the world established Central Statistical analogue in higher dimensional space; plays. The best economists today Agencies such as the Australian Bureau most spatial point process models are are mathematicians able to develop of Statistics. One theory for the rationale intractable whereby moments are not complex models of human economic for setting up these agencies is that known analytically; and the normalising behaviour. governments at time were heavily constant in the likelihood is intractable. In the past, linear regression models influenced by the economic theories Also, time series methods cannot be in economics used a very small number of Keynes. Keynesian economics generalised to spatial processes. of variables with a big emphasis to encourages governments to have a key Ripley advocated a range of solutions transform the model. This dependence central role in managing the economy. such as analytic results for a Uniform on regression has reduced with the If for example, unemployment is high, Poisson process, the method of moments advent of powerful computers and then the government should pump for a stationary process and summarising the development of other parametric/ prime the economy via expansionary point patterns by simple statistics or non-parametric methods. Maximum monetary policy. This level of measuring the distance between points likelihood methods started to be used government control required collection compared to a benchmark (eg. Poisson). in the 70’s and 80’s, and had a big of large quantities of timely and The classical approach for models impact on micro economic models accurate information. Another theory is aims to formulate and fit point process such as labour study models. For that with the introduction of computers, models to point pattern data. The main example, logit and probit models have the ability for the government to collect modelling tool is conditional intensity, been used to develop labour models, and store large volumes of information conditional on probability of getting a such as modelling whether people increased dramatically. point of the process at that location. Point want to work. Maximum likelihood As technology continues to improve process densities used are homogeneous methods have also been useful with at a rapid pace, the cost of acquiring and Poisson, inhomogeneous Poisson and “simultaneity bias”, where one or more storing data decreases, and the demand Strauss Process. The log-linear intensity of the explanatory variables are also for more data increases. Many examples is a large class of models of spatial trends functions of the response variable, eg of large data collections exist such as with covariate effects and an intercept supply and demand. labour market panel studies in the US, interaction. Maximum likelihood for There is an increasing interest in semi- Britain and Australia, supermarket

SSAI Newsletter – February 2004 23 Society Secretaries Branch Reports Continued

Central Council barcode data, stock market data and collected or do they own it? If they sell President: Mr N. Bartlett Secretary: Dr D. Shaw wholesale price of electricity measured in the data to another business, should Email: [email protected] half hour blocks over the last five years. you be entitled to some of the money? The explosion of the incidence of credit 2. What are the implications of the New South Wales President: A/Prof John Rayner risk is encouraging businesses to use accumulation of data as Conglomerates Secretary: Mr Alun Pope data in much greater detail, for example continue to merge and expand? For Email: [email protected] identifying people who have been bad example, if Coles Myer expanded [email protected] credit risks in the past, but who can still their business to the financial services Canberra be profitable. This rapid increase of data market, then the company has the President: Dr Ann Cowling acquisition and use raises a number of potential to link client data from their Secretary: Ms Anna Poskitt interesting questions: two databases (ie matching consumer Email: [email protected] 1. Who owns your data? Do you own data with their financial services data). Victoria the data that your credit company has Peter Ricci President: Dr K. Lipson Secretary: Mr B. Fraser Email: [email protected] Australasian Conferences South Australia President: Dr Alan Branford Secretary: Mr Patrick Lim Statistical approaches to meta-analysis — Three courses for statisticians and Email: [email protected] reviewers in Melbourne and Sydney, March and April 2004 http://www.cochrane.org.au/statistics Western Australia President: Dr Martin Hazelton Workshop on Contaminants and Ecological Risk Assessment Secretary: Ms Jodie Thompson Email: [email protected] 5 – 7 April 2004 — Adelaide – Details: www.clw.csiro.au/conferences/contaminants

Queensland CMA National Research Symposium on Probability and its Applications President: Dr Bronwyn Harch 22 – 23 April 2004 — Canberra Secretary: Dr Petra Kuhnert A meeting to present Chris Heyde with a festschrift on his 65th birthday. Email: [email protected] Info at http://www.maths.anu.edu.au/events/sy2004 Section Chairs 24th International Symposium on Forecasting Statistics in the Medical Sciences 4 – 7 July 2004 — Sydney – Info at: http://www.isf2004.org Peter Howley [email protected] Econometrics Society Australasian Meetings 7 – 9 July 2004 — Melbourne Info at: http://www.monash.edu.au/oce/ESAM04/ Statistics in the Biological Sciences Dr Simon Barry Australian Statistical Conference Email: [email protected] 11 – 16 July 2004 — Cairns, Queensland Survey and Management Contact: Neville Bartlett, [email protected] Dr Robert Clark Email: [email protected] International Biometric Conference 11 – 16 July 2004 — Cairns, Queensland Statistical Education Dr D. Nur Contact: Kaye Basford, [email protected] [email protected] ISBIS 4 – International Symposium on Business and Industrial Statistics Statistical Computing 13 – 16 April 2004 — Cairns, Queensland Associate Professor Kuldeep Kumar Further information at http://www.action-m.com/isbis4/ Email: [email protected]

Industrial Statistics Dr Aloke Phatak Overseas Conferences Email: [email protected]

Young Statisticians TIES 2004 – The International Environmentics Society and ACCURACY 2004: Anna Munday 6th International Symposium on Spatial Accuracy Assessment. Email: [email protected] 28 June – 1 July 2004, Portland, Maine, USA. Bayesian Info at: http://www.ncrs2.fs.fed.us/4801/meetings/ties/default.asp Kerrie Mengersen Email: [email protected] The 6th ICSA International Conference July 21 – 23 2004 — National University of Singapore (NUS), Singapore Further contact details More information can be obtained from the NUS-ICSA 2004 website at: for Society Secretaries http://www.stat.nus.edu.sg/ICSA.htm and Section Chairs can be “Visions of Futuristic Statistical Methodologies” obtained by contacting the 28 – 30 December 2004 — Kandy, Sri Lanka Society on (02) 6249 8266 Conference website: http://www.st.rmit.edu.au/~desilva/conference/slstat.htm

24 SSAI Newsletter – February 2004 SA Branch Christmas Dinner at the Tin Cat Cafe

Peter Ricci and Gary Glonek

Chris and Margaret Brien, Margaret Swincer

Alan James and Graham Wilkinson

SSAI Members among the ISI Debra Partington and David Hirst Top 10! Website of the month Congratulations to Iain Johnstone and Peter Hall, SSAI West Australian Branch member and AStat, John Henstridge, is director members who feature in the of Data Analysis Australia. The company has just turned 15. Their website, Institute for Scientific Information http://www.daa.com.au, describes the company’s achievements since list of researchers with the most 1988 and lists their current activities. These include support for the Young citations between January 1993 Statisticians Section of this Society. Well done John and the team at DAA! and April 2003. www SSAI Newsletter – February 2004 25 The National Undergraduate Statistical Careers Summit NUSCS 2003

On Monday and Tuesday 29- position in 1972. Helen stressed the first statistical consulting companies 30 September 2003 the National importance of ‘connecting’ with in Australia. Then Tim Higgins, an Undergraduate Statistical Careers students and teaching statistics in a actuary and statistician at the ANU, Summit (NUSCS) was held in coherent way. gave an illuminating talk on modelling Canberra. This workshop was part Phil McCloud began the third HECS, doubtful debt and revenue. of a national campaign to increase session, titled Predicting Uncertainty Richard Madden, of the Australian awareness of statistics as a discipline – Farmers and Pharmacy, by talking Institute of Health and Welfare and a career choice. The workshop was about his experiences as a statistician (AIHW), began the seventh session, sponsored by the School of Finance in the pharmaceutical industry, in titled Healthy and Wealthy Statistics, and Applied Statistics (FAS) at the particular Roche since 1997. He Australian National University (ANU), by giving an overview of the gave a very illuminating overview organisation of which he has been the Australian Bureau of Statistics of the strategic life cycle plan of (ABS), and the Statistical Society of director since 1996. He described the a pharmaceutical drug, including values and structure of AIHW, and Australia (SSA). The stimulus for the details of the various clinical trials project came from FAS. pointed out that statisticians need not involved. Then Phil Kokic, of the worry too much about their careers About 40 students were sponsored Australian Bureau of Agriculture on account of their skills being in to attend the workshop. These were and Resource Economics (ABARE), demand almost everywhere. Then selected from over 120 applicants who gave an interesting talk on linking applied from around Australia. The Diane Gibson, Head of the Welfare climate variability with farm financial Division of AIHW, talked about the workshop consisted of a series of talks performance. Phil has developed distinction between survey data and by about 20 leaders in the field from a new economic model which may administration by-product data. She a wide range of sectors, including help predict the impact of climate also discussed evidence-based policy academia, public service, industry, and commodity price variation on and different ways of reporting finance, and private consultancy. The financial performance. following is a brief outline of these statistics. Next Chris Stevenson In the fourth session, titled talks, which took place at the ABS on described his work as a biostatistician Monday and at the ANU on Tuesday. Surveying – The Statistical Landscape, in the Health Division of AIHW. One of More details can be found at http:// David Steele gave an illuminating his projects has to do with determining ecocomm.anu.edu.au/events/nuscs presentation on sample surveys and numbers of bowel cancer patients who related fields, with especial reference Dennis Trewin, the Australian have been cured. Then John Goss, an to the ABS, where he spent 17 years economist in the Resources Division Statistician, opened the Summit by in the Methodology Division, and of AIHW, described a project aimed discussing the importance of statistics the University of Wollongong (UW), at assessing the combined effects of in society, which has been growing where he has been since 1992. David various risk factors for heart disease, because of the masses of databases also described a project of his on the after accounting for interactions and becoming available and the trend Irish Labour Force Survey, which he synergies. towards evidence-based decision was able to redesign with a substantial making. increase in efficiency. Steve Stern of FAS began the eighth Craig McLaren began the first The final talk on Monday was and final session of the Summit, titled session, titled Australia’s Statisticians, by Ross Maller of the ANU, in the Sugar, Safety, Satellites and Epidemics, by describing his work in the Time fifth session, titled Statistics - The by showing how classification trees Series area at the ABS since he joined Finance Frontier. Ross discussed could be used to help clinicians in in 1999. One of his projects has been to the applications of probability and the US determine insulin resistance in develop a method which removes the statistics in the very vigorous field individuals based on other less direct Easter effect from seasonally adjusted of finance, and gave an overview of measurements. His work has been time series. Next, James Chipperfield, some key concepts. presented this year to the American also of the ABS, talked about his Diabetes Association. Then Terry The Summit Dinner was held at work with artificial neural networks O’Neill, Head of FAS, gave a very the Belconnen Premier Inn, where the and computer assisted interviewing entertaining talk on safety measures (CAI). Then Bill Allen, Director of the guest speakers were Terry O’Neill and Michael Martin. In his talk, Michael in motor vehicles, including the use of Household Survey Methodology Unit air bags. Next, Niels Becker from the at ABS, gave several examples of the gave highly entertaining advice on National Centre for Epidemiology and importance of survey sampling and how to receive special treatment when Population Health (NCEPH) at the data analysis. One of his projects is to attending a conference. ANU gave a fascinating presentation optimise the design of the Indigenous Annie Solterbeck of Statistical on the statistical underpinnings Health Survey. Revelations began the sixth session, of epidemics. Niels talked about titled Statistical Consulting – Ask Bronwen Whiting, of the epidemics in general, but his comments the Oracle (Tuesday morning), by Queensland University of Technology had particular relevance given the describing her life as a statistical (QUT), began the second session, recent SARS outbreak. Finally, Simon titled Teaching for Statistics’ Future, consultant. She outlined her typical Barry from the Bureau of Resource by recounting how she came to do a week and discussed some of the skills Sciences spoke dynamically about PhD in Statistics. She talked about the which are important in her line of the statistical problems associated benefits of teaching and academic life, work, such as being very organised. and spoke very enthusiastically about Then Robyn Attewell of Covance with creating accurate land-use maps the usefulness of statistics. Then Helen gave a talk on her experiences as a based on satellite imaging. His talk MacGillivray, also of QUT, described statistician over 20 years. In 1983 she was brilliantly illustrated with maps her experiences of teaching statistics completed a Masters in Statistics at of many parts of Australia. since her first statistical teaching ANU and joined INTSTAT, one of the Borek Puza

26 SSAI Newsletter – February 2004 Report on Biometrics 2003

The biennial conference of the Institute (AMSI) Summer Symposium of large-scale natural experiments in International Biometric Society (IBS) in Bioinformatics was also held at the landscape ecology”), Australasian Region was held at the ANU that week, and Biometrics 2003 • Matt Wand of the University of Australian National University (ANU) in participants were able to make the most of NSW (“Subject-specific curves for early December 2003 and attended by 46 this by attending some of the Microarray longitudinal data: keeping it simple and computable”), biometricians and statisticians. Over the and Experimental Design sessions held • Mick Roberts of Massey University in two-and-a-half days of the conference, on the final day if they were of interest. Auckland, NZ (“A mathematician in At Biometrics 2003, there were 7 invited 25 talks were presented covering the Asia – dining on dengue and supping speakers: themes of ecology, nonparametric mixed with SARS”), models, health, modelling disease spread, • David Lindenmayer and Ross • Annette O’Connor of the Office of the microarrays and various applications. Cunningham of the ANU (“Issues in the Chief Veterinary Officer in Canberra The Australian Mathematical Sciences design, implementation and analysis (“The role of infectious disease spread models in Australia’s exotic animal disease response preparedness”), • David Baird of AgResearch, NZ (“Design, normalisation and quality control for two colour cDNA microarray experiments”), and • Harold Henderson of AgResearch, NZ (“Dynamic graphics for microarray data”).

John Reynolds presents Carole Wright Nick Nicholls, Annette O’Connor and Mick Roberts. with her prize.

Roger Littlejohn, Bob Forrester and Melissa Dobbie.

SSAI Newsletter – February 2004 27 All gave well-polished talks and The AGM of the IBS Australasian Honours/4th year scholarships starting represented a range of backgrounds Region was held during the conference in 2004 and further details about this will and subject areas within the Biometrics and the following office bearers were be announced shortly. domain. voted in: Being a boutique conference (by Four students gave presentations at President: John Reynolds (Peter default), there was adequate opportunity the conference and for their efforts, the MacCallum Cancer Center) for conversation with new and old faces society provides free membership for Vice President: Ann Cowling (Australian and certainly no awkward decisions one year to each student. However, there National University) about which sessions to attend, as there were also monetary awards on offer at were no parallel sessions. Congratulations Treasurer: Warren Muller (CSIRO) with this conference. Carole Wright of the are deserved to the Local Organising and help from Brian Niven (University of Program Committees, both chaired by University of Waikato, Hamilton, NZ Otago, NZ) took out the Platinum award for her talk Ann Cowling of ANU, for organising a Secretary on “Quick construction of resolvable row- : Dave Saville (AgResearch, NZ) worthwhile and successful conference at column designs” and Marie Forrester of Biometric Bulletin Correspondent short notice. Queensland University of Technology, (Australasian region): Melissa Dobbie Further details about the conference, Brisbane took out the Gold award for her (CSIRO) such as abstracts of all papers, copies talk on “Statistical and stochastic epidemic At the AGM, the decision was made to of most contributed and invited models of MRSA occurrence within an offer some financial support for student presentations, and selected photos taken intensive care unit given incomplete members to attend the upcoming IBC/ of participants, are available at http:// observations”. Congratulations to all ASC2004 meeting in Cairns in July 2004, www.maths.anu.edu.au/events/IBS02/ four speakers on the quality of their so watch this space for more details! index/html presentations. The region is also planning to award Melissa Dobbie, CSIRO QLD

Warren Muller, Peter Johnstone and Ky Mathews.

Helen Nicol, Richard Morton and David Butler.

28 SSAI Newsletter – February 2004