Our History

In the early 1990s, Stephen Morris was involved in providing ‘refresher’ courses in statistics to medical school researchers. Typically they had all taken statistics as part of their training, but unfortunately, they had gained no real understanding of how it worked and felt that it would always remain something of a mystery. Like most people, they just wanted to know how to use statistics in their work: they were not at all interested in Statistics as a subject, or in complex mathematical proofs of how techniques worked. Their main concern was that their research should not be jeopardised by bad statistics.

The breakthrough came with choosing tests. This was an area of particular confusion: course participants seemed to have no way of judging which tests should be used with different configurations of data. Stephen evolved a technique for teaching them some ‘rules of thumb’ which would carry them forward in the majority of cases. This was based on pattern recognition; the natural layout of data on the page can be used to distinguish between configurations of data requiring different families of tests. The idea developed of computerising this aspect of the course and adding a quiz. This was the first section of what was to become Statistics for the Terrified; improved and refined, it is still there as the section ‘How to choose a test’.

Once we had begun, ideas for the presentation of data in other graphical ways came naturally. The section on regression was one of the next parts to be developed, and we rapidly realised that the interactive graph was a popular and effective way of teaching statistics. The sight of pairs of students racing each other to complete line-fitting challenges successfully was a revelation: we had never seen such enthusiasm for statistics in the classroom before. Another interesting side-effect of these exercises was that by moving the points on the graph and watching the changing position of the ‘line of best fit’, students were incidentally gaining an intuitive understanding of the disproportionate effect that outliers can have. In v3.0, therefore, we added a short section on this at the end of the regression module.

This is a good example of how we have always responded to the feedback we have received on the way the software has performed in practice. Sometimes there have been unexpected benefits (as with the outliers) and sometimes new sections have not gone as well as we had hoped. With each version, new material has been added (on non-parametric statistics, for example), and existing material and the interface improved. Stephen’s teaching experience enabled us from the start to deal with areas of common confusion, and we are always on the lookout for these.

As the material was developed it was quickly adopted for use by students at undergraduate level in a wide variety of disciplines, from medicine to archaeology. Since the first version, students have reported increased confidence in their ability to deal with statistics, and increased understanding of other teaching and of materials such as textbooks and sample analyses. Having gained a commonsense understanding of what statistics in general, and individual techniques in particular, are trying to acheive, they find it much easier to follow the traditional mathematical explanations.