Report on the Inaugural Meeting of ASSUME, 27th March 1996

by Neville Hunt, Coventry University

Having been rearranged due to snow, the RSS Spreadsheet Forum finally took place on 27th March at the Society's Centre for Statistical Education at Nottingham University. Centre Director Anne Hawkins welcomed the 21 participants on behalf of the Society and the Centre, then it was over to Neville Hunt (Coventry University) to introduce what he saw as the key issues for debate.

Using a live Excel demonstration to show examples, Neville was enthusiastic about the wide variety of spreadsheet activities that could be used in teaching statistical ideas. These included "Blue Peter" spreadsheets ("here is one that I produced earlier"!), DIY spreadsheets (constructed by students themselves), macro-driven simulations, and templates for basic statistical analyses. He was less happy about using spreadsheets for "doing" statistics. Using Excel's Pivot Table and Trendline facilities as examples, Neville showed that these potentially excellent features were marred by annoying details and crass errors (including an R-squared of -198%!). To summarise, Excel offered a mixture of the good, the bad and the ugly.

There followed an hour of animated discussion in small groups, followed by a report-back time chaired by James Currall (CTI Centre, Glasgow). Each group shared its views on what was good, what was downright wrong, what could be improved (the "wish list"), and what was missing in Excel's statistical capabilities. Discussion continued over a well-earned buffet lunch.

The afternoon programme, chaired by Glyn Davis (University of Teesside) comprised four fascinating 20 minute presentations. First to go was John Cooney (University of Leeds Medical School), who demonstrated his ASTUTE suite of statistical add-ins for Excel. The add-ins encompass a wide range of statistical techniques, including non-parametric tests. An additional module is available for specialist medical applications. The ASTUTE programs use in-house fully-tested algorithms rather than Excel's inbuilt functions. One impressive feature was the facility to customise the output.

Mike Talbot of BioSS (Edinburgh) described how he and his colleagues had, reluctantly at first, been persuaded to use Excel for three different purposes: training in basic statistics for research scientists; developing add-ins for specialist techniques; providing an interface to modelling programs. Mike explained that it was difficult to persuade scientists to use the best methods if this meant they must go outside their preferred working environment (Excel). Excel had been used to provide a seamless interface with a Fortran program to perform REML analysis, to avoid the need to use a more advanced package such as Genstat.

The next presentation was given by Chris Ricketts (University of Plymouth) on resampling techniques. Chris demonstrated how easy it was in Excel to carry out the sort of randomisation tests which lie behind most modern non-parametric tests. Having set up a spreadsheet to test for differences between means from two small non-Normal samples, it was easily modified to test differences between medians, or even standard deviations (try deriving the sampling distribution for that!). Bootstrap confidence intervals were equally easy to produce. Experience with students suggested that they found this approach far more comprehensible than the traditional approach to parametric testing.

The final presentation, keenly awaited, was given by Paul Barnwell representing Microsoft. Paul welcomed the opportunity toUse of uninitialized value in concatenation (.) or string at E:\listplex\SYSTEM\SCRIPTS\filearea.cgi line 455, line 86. get feedback from statistics specialists. Acknowledging our lengthy "wish list", Paul reminded us that Excel was aimed primarily at the business and financial sectors. He explained that Microsoft support was essentially phone-based and the company could not respond in detail to letters and faxes. However, a huge "knowledge base" of technical support is available at the Microsoft WWW site (http://www.microsoft.com) containing literally thousands of tips, workarounds and macros generated by users' enquiries. Requests for new facilities or modifications can also be emailed to the Microsoft wish alias (ukoffwsh@microsoft.com). With a new version of Microsoft Office in the pipeline, there is a possibility that some statistical expertise might be sought for Beta testing.

In a concluding discussion before tea it was decided to establish ASSUME (Association of Statistical Specialists Using Microsoft Excel). Annual meetings will be organised by Neville Hunt. Roger Silk (Sunderland) will edit WWW pages for the users' group, which will be accessible from the CTI Centre site at Glasgow. Members of ASSUME will communicate via a closed email list initially. Anyone wishing to be added to the mailing list should contact Neville Hunt (N.Hunt@coventry.ac.uk). Links with the Society's Business and Industrial Section will also be fostered. It was generally agreed that a profitable day had been enjoyed - thanks to all who participated.