Fishing in the Bay

(Posted on behalf of Ken Russell). The Mathematics and Statistics in Industry Study Group Conference (MISG) for 2007 will be held at the University of Wollongong, from 5th to 9th February 2007, with A/Prof. Tim Marchant as director.

Please note that the Group does now include Statistics in its name; the awareness that statisticians are a vital part of the week’s proceedings has been growing steadily in recent years. More details follow below. Please consider attending.

The MISG conference is an industry clinic where industry partners have their real world problems solved by the delegates (who normally number around 100-150). This type of conference has a wide and successful history around the world. See HERE and also below for a list of current and past meetings and details of problems solved.

The industry partners present their problems to the Group on the Monday. During the week, the MISG attendees make initial attempts at the problems, and present an interim solution on the Friday morning. Over the rest of the year, further work is done on the problem, and the results are published at the beginning of the following year. Each problem is allocated two senior researchers, and a postgraduate student, as moderators. The other MISG attendees may contribute to one or more problems, but most seem to adopt just one.

In recent years, there have been many problems of a statistical nature at MISG, and 2007 will be no exception. Specific details of some of the industry problems for 2007 are still being finalised but we expect three or four problems to require statistical modelling, financial mathematics or involve operations research. As the problems are finalised they will be posted on the web-site, www.misg.math.uow.edu.au.

There is no attendance fee. Tea and coffee, and one or two lunches, are provided. The cost to attendees is in getting to Wollongong, and in accommodation. Information on accommodation and transport is on the web-site.

Who should attend MISG 2007? Postgraduate students, junior and senior researchers, … anyone with an interest in practical problems. We expect groups from Statistics New Zealand and the Australian Bureau of Statistics to attend. What will you get from it? Real experience, the chance to meet like-minded colleagues, and an appreciation of how mathematics and statistics can be applied (I’ve learnt a lot at the last two MISGs). One other reason to attend: we’ve invested a lot of effort in recent years convincing applied mathematicians that statisticians do play an important (and *not* subsidiary) role in problem-solving. We’d like to reinforce that.

Here are a couple of proejcts from the MISG meeting held in 2006 meeting to give you teh flavour of the kinds of statistical projects that might arise.

The first, entitled “Multi-variable relationships in a batch annealing process”, was proposed by New Zealand Steel. They collect readings on many variables as their cold rolled steel is manufactured from hot rolled coil. They wanted a set of equations that would let them predict the values of four output variables in terms of the values of many input variables or factors.

The client for the second problem was the Sustainable Soil Management Promotion Group, led by soil and agricultural scientists from Crop & Food Research and Environment Canterbury, New Zealand. The title was “Predicting the effects of agricultural land management change on soil quality and productivity”.

Scientists had developed a simple model designed to allow Canterbury farmers and resource managers to predict changes in soil quality and associated productivity based on crop and soil management information on arable cropping farms. The aim of the current project was to expand the model to cover a wider range of land uses across at least six different regions within New Zealand where soil types, climate, and agricultural management practices differ. The client wanted a model that could be used by farmers, agricultural consultants and resource managers to predict the likely consequences of changing management practices on soil conditions and associated productivity under these different land uses.

Both problems essentially involved multiple regression. Alternative analysis methods could be used if the participants wanted, but it had to be borne in mind that the ultimate aim was to give the clients something that they could use and understand. As in all real problems, the data sets required cleaning and one could not assume that every variable was exactly as described or consistently measured. Moreover, the data came from observational studies, not designed experiments. These problems were a far cry from the analysis of a small and carefully-cleaned dataset, with nicely defined and measured variables, and represented an excellent for junior - and senior - statisticians to develop further their problem-solving skills.

You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.