Using Statistics for the Terrified as a teaching resource in Universities and Colleges

A good teacher with plenty of time is always the best solution. Currently, however, teachers have to deal with more and more students in less and less time, and the more high-quality resources they can employ in their teaching, the better the results will be.

Teaching statistics is a problem. It is a mathematical discipline, and it is seen as such; but at the same time it is an essential research tool in even the most non-mathematical disciplines. Unfortunately the traditional courses are based on the mathematical foundations of statistics, and all too often students are shocked and dismayed by this. They hate it because they feel unequal to the maths content – and the teachers hate it because they don’t want to have to teach maths. They want to teach their own subject and the maths is getting in the way.

Statistics for the Terrified is a non-mathematical course designed to provide students with a fundamental grasp of the principles underlying statistics and statistical analysis, so that they can carry out any research projects and critical appraisal required of them, and approach other materials (such as textbooks) with increased confidence.

Statistics for the Terrified reformulates the presentation of ideas away from the traditional mathematical approach and into straightforward puzzles and pattern recognition. Pattern recognition is our most basic and most practised learning skill, which we have been using since childhood. The mathematical baggage is put to one side, so that there are few prerequisites.

By providing the foundations of statistical understanding in a resource which can be accessed outside contact time, the teacher doesn’t have to explain the mathematics, and is able to concentrate instead on the applications of statistics in their own field. Statistics for the Terrified is good for the teacher because it makes the student want to learn: it is interactive, immediate, fun and yet sophisticated. It provides a sense of real achievement and a true understanding of the mechanisms of research.

As well as its suitability for the standard undergraduate statistics course taken as part of another subject degree, this tutorial is ideal for those teaching students who require critical appraisal or research skills, and will also enable those undertaking research projects to select the most appropriate tests and techniques. Should they need to understand more sophisticated statistics, they will have gained insight which will carry them forward.

A typical classroom experience might be as follows:

In a classroom session devoted to the technique of regression (or linefitting) the teacher can use the software (projected onto a large screen) to graphically demonstrate the purpose and uses of regression. The tutorial contains animated examples for the class to see.

The three statistical summaries by which the ‘fit’ of the line to the data is described (gradient/slope, intercept and correlation) are then introduced by the teacher following the flow of Statistics for the Terrified. The next section of the software is designed for the students to spend a period of time exploring these, perhaps five or ten minutes, using the dynamic graph. Data is changed by dragging points on their screens; the line of best fit is recalculated and redrawn automatically each time a point is moved. They can then observe how the visual fit of the line changes and the effect on the three values. This provides the beginnings of an intuitive understanding of the relationships between the data, the line of best fit, and these three key parameters.

The teacher then moves the students on to a series of exercises or ‘datagames’. The tutorial provides randomly generated data sets in the dynamic graph, and the students are challenged to change the data in order to reach randomly generated target values for each of the gradient, intercept and correlation in turn.

These exercises are a very powerful method of giving users a firm grasp of what the values actually represent. For example, many students start out with the impression that slope and correlation are the same: but working through these challenges they can get a high correlation with a particular slope and then a low correlation with exactly the same slope. Slope and correlation are not the same thing; it is obvious now.

The next set of challenges is more difficult but more realistic: the target is a combination of values, for example a particular slope with a particular correlation – the targets are again generated randomly.

Because the targets of all these exercises are different each time, the user cannot learn to succeed in them without gaining some real understanding. Exercises like this are surprisingly rich: the regression exercise lays the ground for an exploration of outliers, as students directly experience the disproportionate effect of moving the outlying points.

After formal teaching sessions, these exercises are available to the individual student for consolidation of their learning, on the institutional network or open-access PCs.