Bayes on the Beach 2011

6th & 7th October
Vibe Hotel, Surfers Paradise

• Registration
• Abstract for International keynote speaker: Sudipto Banerjee, University of Minnesota
• Abstract for Australian keynote speaker: Edward Cripps, University of Western Australia
• About the Keynote Speakers
• Introduction to Hierarchial Modelling for Spatial Data Workshop
• Workshop Schedule
• Organising Committee

The conference will provide a forum for discussion on developments and applications of Bayesian statistics. The International keynote speaker is Professor Sudipto Banerjee from the University of Minnesota and the Australian keynote speaker is Assistant Professor Edward Cripps from the University of Western Australia. The format includes seminars, contributed sessions, a poster session, tutorials and workshops.

Registration will open on the 1st September 2011.
Students Registration fee $200.00
Regular Registration fee $400.00

Please contact Vibe Hotel Gold Coast to book accommodation
http://www.vibehotels.com.au/default.asp?page=/vibe-locations/gold-coast-hotels/vibe-hotel-gold-coast

Call for abstract for Poster session and Contributed session closes Thursday 15th September. 200 words abstract only and please indicate if you wish the abstract to be considered for a poster or a contributed talk.

All abstracts should be sent to Dow at email address dow.jaemjamrat@qut.edu.au.

In addition to the conference there will be a one day workshop by Sudipto Banerjee held at QUT.

Registration

To download a registration form, for Bayes on the Beach please click here.

To download a registration form, for Introduction to Hierarchical Modelling for Spatial Data Workshop please click here.

Please return this form by mail or by fax by 30 September 2011 to:
The Executive Officer, SSAI, PO Box 213, Belconnen, ACT 2616
E-mail: eo@statsoc.org.au
Fax: (02) 6251 0204
Ph: (02) 6251 3647
Statistical Society of Australia Inc, ABN: 82 853 491 081

Please also email dow.jaemjamrat@qut.edu.au to let the organisers know you are attending and to advise of any special dietary requirements which you may have.

International keynote speaker: Sudipto Banerjee, University of Minnesota

Title: Computationally feasible hierarchical modelling strategies for large spatial datasets

Abstract:
Large point referenced datasets are common in the environmental and natural sciences. The computational burden in fitting large spatial datasets undermines estimation of Bayesian models. We explore several improvements in low-rank and other scalable spatial process models including reduction of biases and process-based modelling of ``centres'' or ``knots'' that determine optimal subspaces for data projection. We also consider alternate strategies for handling massive spatial datasets. One approach concerns developing process-based super-population models and developing Bayesian finite-population sampling techniques for spatial data. We also explore model-based simultaneous dimension-reduction in space, time and the number of variables. Flexible and rich hierarchical modelling applications in forestry are demonstrated.
Keywords: Bayesian modelling, Low-rank Gaussian processes, Hierarchical modelling, Markov chain Monte Carlo, Spatial data, Spatial super-populations
[back to top]

Australian keynote speaker: Edward Cripps, University of Western Australia

Title: Mixture of random effects for individual's learning behaviour

Abstract:
We present a dynamic mixture of random effects model applied to a current topic in the psychology literature. In psychology, the implicit theory of abilities proposes that individuals are classified as one of two groups:entity theorists who believe ability is innate and incremental theorists who believe ability is anaquired set of skills. Entity theorists are more likely to interpret failure as evidenceof a lack of ability and doubt their future capacity to learn the task. Incremental theorists are more likely to interpret failures as part of a learning strategy, potentially leading to recovery over time. The hypothesis is that learning performances of entity theorists are more prone to downward ``spirals'' than incremental theorists. The Accelerated Learning Laboratory, Melbourne Business School, has conducted experiments in which individuals from both groups were subjected to repeated tasks and their performances evaluated. To assess the hypothesis we model the performance of an individual as a function of that individual's personality self-classification using a time-varying mixture of potentially two constrained random effects models, one before spiralling behaviour begins and another after. Performance is modelled dynamically by allowing for the commencement of the spiral to be a function of time and to vary with individuals. So for each individual we average over another class of models which are the possible locations of the spiral, and performance is predicted by weighting all possible models by their posterior probability.
[back to top]

Sudipto Banerjee received an MS and PhD in statistics from the University of Connecticut. Prior to this he received a B.Sc. (Honours) from Presidency College and an M.STAT from the Indian Statistical Institute, both in Calcutta (now called Kolkata), India. He is currently a tenured Professor of Biostatistics in the School of Public Health, University of Minnesota Twin Cities. His research focuses upon statistical modelling and analysis of geographically referenced datasets, Bayesian statistics (theory, methods and applications), statistical computing/software and the melding of numerical/physical models with observational field data. He has published over seventy peer-reviewed journal articles, several book chapters and has co-authored a book titled "Hierarchical Modelling and Analysis for Spatial Data". He has overseen the development of several Bayesian software packages within the R statistical framework. In 2009 he received the Abdel El Sharaawi Award from the The International Environmetrics Society -- accorded to a young investigator (below the age of 40) every year who has made outstanding contributions to the field of Environmetrics. In 2011 he was honoured with the Mortimer Spiegelman Award, given since 1970 by the American Public Health Association to an outstanding public health statistician below the age of forty. Sudipto is also an elected member of the International Statistical Institute.

http://www.sph.umn.edu/biostatistics/ourfaculty/faculty/baner009
[back to top]

Edward Cripps is Assistant Professor at the School of Mathematics and Statistics, University of Western Australia.
Edward completed a Bachelor of Economics (1995) with the School of Economics,University of Western Australia. He then went on to achieve 1st Class Honours, with the Faculty of Science (1999), at the School of Mathematics, with the University of New South Wales.

Edward Cripps Completed his PhD in Sept 2005 with the School of Mathematics, at the University of New South Wales.
Thesis was on "Model uncertainty in univariate and multivariate Gaussian linear regression." (Edward Cripps advisors Dr David Nott and Professor Robert Kohn).

His research include computational statistics, Bayesian methods and applications, longitudinal data, model averaging and mixture models.
[back to top]

Hierarchical Modelling for Spatial Data Workshop Details

By: Prof. Sudipto Banerjee (U. of Minnesota)
Date: Wednesday 5th October
Room: S405
Building: S Block
Campus: Gardens Point Campus, Queensland University of Technology

Registration:
Students $80.00
Conference attendees $150.00
Workshop Only $300.00

Please contact Dow for any further details at dow.jaemjamrat@qut.edu.au

Recent advances in Geographical Information Systems (GIS) and Global Positioning Systems (GPS) enable accurate geo-coding of locations where scientific data are collected. This has encouraged collection of spatial-temporal datasets in many fields and has generated considerable interest in statistical modeling for time and location-referenced data. Bayesian hierarchical models for the analysis of spatial data continue to be popular because of their ability to offer direct inference and the incorporation of prior information in inference. Until recently, practical Bayesian data analysis was somewhat encumbered by the limited availability of user-friendly Bayesian software packages for spatial data. Win/OpenBUGS, the popular Bayesian software package was efficient only for a certain special class of spatial models for regionally-aggregated data, but could not handle much more beyond that. The state of affairs has dramatically changed over the last few years with the advent of several R packages that not only offer user-friendly interfaces to Bayesian spatial models, but also interact with Geographical Information Systems (GIS). This course will offer a hands-on opportunity to explore the use of OpenBUGS, the leading Bayesian software package, as well as several spatial packages in R for geo-coded data. The computing demonstrations will encompass exploratory spatial data analysis as well as estimation of statistical models with practical data sets in public health and the environmental sciences. Specific topics that will be covered include: geostatistical modeling, spatial risk assesment, disease mapping, spatial linear and generalized linear models, space-time modeling, model diagnostics and model assessment and will especially focus upon practical computations and implementations. Prerequisites include a basic course in mathematical statistics and linear regression models at the Master's level. Familiarity with the basics of Bayesian inference will be useful, but will not be assumed for the course.
Specific topics that will be covered include: geostatistical modelling, spatial linear regression, generalized linear models, models for areally aggregated data and disease mapping.

We will offer a hands-on opportunity to explore the use of WinBUGS, the leading Bayesian software package, as well as several spatial packages in R for spatial geocoded and areal data. The computing demonstrations will subsume exploratory spatial data analysis as well as estimation of statistical models with practical data sets in public health and the environmental sciences.

The following are useful text books for Bayesian statistics and hierarchical models for spatial data analysis:

• Banerjee, S., Carlin, B.P. and Gelfand, A.E. (2004). Hierarchical Modeling and Analysis for Spatial Data. Publisher: CRC/Chapman and Hall.
• Diggle, P.J. and Ribeiro Jr., P.J. (2007). Model-based Geostatistics. Publisher: Springer.
• Waller, L. and Gotway, C. (2004). Applied Spatial Statistics for Public Health Data. Publishers: John Wiley and Sons.
• Carlin, B.P. and Louis, T.A. (2000).Bayes and Empirical Bayes Methods for Data Analysis. Second Edition. Publisher: CRC/Chapman and Hall.
• Diggle, P., Fuentes, M., Gelfand, A.E. and Guttorp, P. (2010). Handbook of Spatial Statistics. Publisher: CRC/Taylor and Francis.
• Gelman, A., Carlin, J.B., Stern, H.S. and Rubin, D.B. (2004). Bayesian Data Analysis. Second Edition. Publisher: CRC/Chapman and Hall.
• Dalgaard, P. (2002). Introductory Statistics with R.
• Faraway, J.J. (2005). Linear Models with R. Publisher: CRC/Chapman and Hall.
• Lee, P. M. (2004). Bayesian Statistics Publisher: Hodder Arnold.
• Venables, W.N., Smith, D.M. and the R Development Core Team (2002). An Introduction to R: Revised and Updated.
• The web sites for softwares:
• WinBUGS or OpenBUGS
• You can download the new registration key for WinBUGS from HERE. NOTE THAT YOU DO NOT REQUIRE ANY REGISTRATION KEY FOR OPENBUGS.
• R.

[back to top]

Workshop Schedule

[back to top]

Organising Committee

• Professor Kerrie Mengersen (co-Chair)
• Dr Clair Alston (co-Chair)
• Dow Jaemjamrat

For further information please contact Dow Jaemjaret at dow.jaemjamrat@qut.edu.au or telephone +61 7 3138 2063