## Research InterestsBiometrics; mixture models; model selection; statistical computing; robust estimation and diagnostics. Dr Murray Jorgensen BSc(Hons) Room:G2.19 Phone +64 7 838 4773 Fax +64 7 838 4155 |

I joined the University of Waikato in 1986 coming from a statistical consulting position with the Ministry of Agriculture and Fisheries in Wellington. My previous academic positions were with the University of Botswana in Southern Africa and Wilfrid Laurier University in Waterloo, Ontario, Canada. My undergraduate education was at the Universities of Auckland and Canterbury and I have a PhD in Mathematics from the University of British Columbia.

My current research is in the area of statistical computing, especially in applications of the EM algorithm. Mixtures of statistical models is an example of such an application and can be quite useful as many real data sets can profitably be regarded as mixtures from several sources. An example is data from the New Zealand census which can be obtained for nearly 2000 small areas but which becomes more understandable if a good way of grouping these areas into regions with similar data can be found. I have also been interested in implementing the Minimum Message Length approach to estimation and model selection using the EM algorithm. Recently I have been looking at the different age/sex patterns of the New Zealand population. This leads to an interesting division of New Zealand into regions of particular age/sex compositions

The Multimix web site contains a Fortran 77 program for fitting a class of models that includes latent class models and finite mixtures of multivariate normal distributions. An allocation rule based on the fitted model can be used as a form of cluster analysis. Documentation and a background paper is included at the site.

**Rohan Maheswaran** is working on a method for making statistical models less sesitive
to dubious data using mixtures of statistical models.

**Paul Taylor** is looking at the use of statistical methods to develop indices of
fish abundance using data collected from pilots of fish spotting aircraft.

I have a great love for general strategies for analysing data, rather than for elaborate approaches that apply only to a single set of circumstances. Statistical Science as it develops has identified much common structure in the data analysis problems of many disciplines. This is quite exciting because it means the tools and techniques developed to serve the needs of well-funded subjects can often be applied to other subject areas.

I am a proud father of two sons:

**Andrew** is a Philosopher currently
at University College, Dublin. He is very interested in things that do not appear
to exist in any real sense, but that we seem to need to talk about.
(God knows how he got started on that!)

**Mathew** is owner and manager of *Forté* - a nite club and bar on Fort Lane
in downtown Auckland.

Courses that I expect to teach in 2008 are:

STAT522A Statistical Inference.

This is a graduate course developing the
theory of modern statistics based on the ``Likelihood Function''. The focus
will be on those parts of the theory which are encountered in the practice of
Statistics.

STAT323A Design and Analysis of Experiments and Surveys.

I will be discussing some of the classical designs for experiments and surveys
which serve as the basis for more complex designs.

STAT121A Intoduction to Statistics.

I'm back in front of this big class this year
after a break of a few years. I hope I haven't lost the knack of show business!.

I hope to take leave in the B semester, but this is not yet certain.

Jorgensen, Murray A. & McLachlan, Geoffrey J. (2008)
Wallace's Approach to Unsupervised Learning: The Snob Program
*The Computer Journal* Advance Access published on January 27, 2008.
doi:10.1093/comjnl/bxm121

Zuma, K., Jorgensen, M., Lurie, M. (2006)
Analysis of interval-censored data from circular migrant and non-migrant sexual
partnerships using the EM algorithm. *Statistics in Medicine* 26, 309-319.

Jorgensen, M.A. (2005) Minimum message length estimation using EM methods: a case study.
*Computational Statistics & Data Analysis* 49, 147-167.

Jorgensen, M.A. (2004) Using multinomial mixture models to cluster internet traffic. *Aust. N.Z. J. Stat.* 46(2), 205-218

Reed, W.J. and Jorgensen, M.A. (2004) The double Pareto-lognormal distribution - A new parametric model for size distributions *Communications in Statistics: Theory and Methods* 33(8), 1733-1753.

Hunt, L.A. & Jorgensen, M.A., (2003). ‘Mixture model clustering for mixed
data with missing information’. *Computational Statistics & Data Analysis*,
41, 429-440.

Expectation-maximization algorithm, article in Encyclopedia of Environmetrics, Wiley, New York, 2001.

Iteratively reweighted least squares, article in Encyclopedia of Environmetrics, Wiley, New York, 2001

Robust regression, article in Encyclopedia of Environmetrics, Wiley, New York, 2001

Method Comparison via Single Factor Analysis. *Proceedings of the 16th
International Workshop on Statistical Modelling*, 243-250, Odense, 2001.

Clustering via Mixture Models: some issues. *Proceedings of the 10th
International Symposium on Applied Stochastic Models and Data Analysis* 2,
585-590, Compiegne, 2001.

A dynamic EM algorithm for estimating mixture proportions. *Statistics and
Computing* 9, 299-302, 1999.

Mixture Model Clustering using the MULTIMIX program (with L.A. Hunt) *Austral.
& New Zealand J Statistics* 41, 153-171, 1999.

Model-Robust parameter dispersions for iteratively reweighted least squares.
*Communications in Statistics: Theory and Methods* 28(8), 1903-1919, 1999.

A significance test for empty corners in scatter diagrams (with W.E.
Bardsley, P. Alpert and T. Ben-Gai) *Journal of Hydrology*219, 1-6, 1999.

Data Mining (with R. Gentleman) *Chance* 11, 42, 34-39, 1998.

Mixture Model Clustering of Data Sets with Categorical and Continuous
Variables (with L. A. Hunt) *Proceedings of the Conference, ISIS 96, Australia*
1996, p375-384.

Tail functions and iterative weights in binary regression. *American
Statistician* 48, 1994.

Influence functions for iteratively defined statistics *Biometrika* 80,
253-265, 1993.

Mixed model discrete regression (with J. Zhaorong and C.A. McGilchrist),
*Biometrical Journal* 34, 691-700, 1992.

Influence based diagnostics for finite mixture models *Biometrics* 46,
1047-1058, 1990.

Fitting nonlinear models: Keep it simple. *New Zealand Statistician* 24, 36-42,
1989.

Jackknifing, fixed points of iterations *Biometrika* 74, 207-211, 1987.