R.D. Stern and E.F. Allan
Statistical Services Centre, University of Reading, Harry Pitt Building
Whiteknights Road, Reading, UK
E-mail: r.d.stern@reading.ac.uk and e.f.allan@reading.ac.uk
Computers are becoming accessible to students, at least at postgraduate level, in universities in Africa and we show how this development can be exploited to support the teaching of statistics.
We do need to consider changing the way we teach statistics. Currently research data are often poorly collected and incompletely analysed. One reason is the lack of understanding of statistical ideas by researchers. This is despite training in statistics. Trainers must therefore accept some responsibility for this lack of understanding. We believe that one way to make the major changes that are required in teaching is through imaginative use of computers within statistics courses.
Once computers are available it is temptingly easy to continue with the same course as before, using computers merely as a teaching aid to help cover the existing materials. This is not sufficient, for reasons we have explained elsewhere (Stern and Allan, Allan and Stern, this proceedings).
At Reading we currently use Excel, Minitab and Genstat on the statistics courses for agriculture students. Instat has also been freely available as a DOS-based package from the Statistical Services Centre since 1994. Recently we have been able to upgrade it, by adding a Windows front end. We are, at the same time, incorporating some of our statistical games and our Good Practice Guides (see Allan and Stern, this proceedings) that are part of our broadened approach to the teaching of statistics at Reading. We anticipate that the first Windows version of Instat will be ready by the end of the first half of 2000.
We believe that Instat+, in its new form, may be useful for trainers of statistics, especially within universities. We envisage that trainers may often want to use more than one software package on a course, and so this package is intended to complement, rather than compete with, other statistical software used in teaching.
Our objective here is to test this belief, by demonstrating the new Instat+, and to introduce general ideas on how to incorporate computers effectively into the teaching of statistics.
The core of Instat+ is a simple statistics package that has been tried and tested on PC compatible computers since the 1980s. It is a beginner's package and therefore includes all the statistical features that would support basic training courses in statistics. It has always been easy to use and is now even simpler, being in Windows. We have tried to support good statistical practice within Instat and have maintained this view in the development of Instat+. Hence Instat+ is particularly strong in simple data manipulations (a missing element in many training courses) and in basic statistical methods. These basic methods include descriptive tools, graphics for data exploration, one sample and two sample inference, good analysis of variance and simple and multiple regression.
The new Instat+ imports data from a wide range of other programs including Excel and Mstat. As was the case for Instat, Instat+ can export data to other packages, including Genstat, SAS and SPSS, when analyses are required that exceed Instat's capabilities.
Students need to understand how analyses can be tailored to the specific objectives of the research. Instat+ caters for this by including such elements as contrasts in the analysis of experimental data.
Students also need to be exposed to a range of advanced statistical methods that correspond to common situations with real data. Thus, Instat+ includes facilities for handling unbalanced experimental designs. It also has a module for fitting log-linear models to count data-a modern method that extends the simple chi-square test to more complex data structures, and shows how advanced methods fit into a general modelling framework.
We have stated that Instat+ complements, rather than competes with, the standard software packages. One role for Instat+ may be to introduce Excel users to the scope of a statistics package for situations when Excel is insufficient. Courses can then move on to use a more powerful statistics packages if the needs exceed Instat+'s capabilities. For example, Instat+ does not have facilities for multivariate analysis (except multiple regression), nor are its facilities for the analysis of unbalanced and non-normal data as extensive as those in more powerful packages such as Genstat and SAS.
A major component of Instat+ is a facility to analyse climatic data in ways that are of direct relevance to their use in agriculture. These facilities for processing climatic data are of some value to agricultural researchers generally and they provide an application area that is of common interest to many students.
Instat+'s documentation, which is over 1000 pages, is included within the Instat+ HELP system. This system also includes example datasets to illustrate different methods of analysis.
We also plan to include the Good Practice Guides within the Instat+ HELP system.
The computer is a key tool in making training in statistics more interesting and relevant to students and researchers. Our objective, in demonstrating Instat+ within this workshop, is to provide encouragement to trainers in statistics who wish to broaden the way they currently teach statistics.
Interested trainers of statistics may want to only use selected components of the Instat+ package. For instance, a university department may already be using Minitab successfully for hands-on computer practical work on their statistics courses, but might want to explore the use of our statistical games as a way of demonstrating good experimental design practice. Others may have no access to statistical software because of financial constraints. For them we hope that the fact that Instat+ will be freely downloadable-as well as encouraging good statistical practice-will help them to make their statistics training more relevant.
