The Data Game: Learning Statistics
Steven Morris and Jill Szuscikiwicz
Statistics can be a problematic subject to learn, especially for students with a non-mathematical background. This article outlines the development of the software package "Statistics for the Terrified", which aims to overcome these deficiencies. The software adopts a new approach to statistical comprehension.

Overview
Overview

Introduction

Learning statistics is a perennial problem for students and research workers from non-mathematical backgrounds. However, the teaching of statistics throughout higher education assumes a high d egree of mathematical competence, even when the students are from non-mathematical disciplines. This is a major reason for students’ perceived lack of statistical judgement; clearly a different approach is called for. With imagination, it is possible to convert difficult statistical problems into simpler problems of pattern recognition, and with computerisation, the extra ingredient of interactivity can be added, enabling the student to manipulate the raw data whilst watching the patterns change. This b uilds up an intuitive understanding of how statistics works. In effect, statistics is turned into a game, the data game.
Overview

The Problem

 Imagine learning to play chess, with a set of principles and examples in a book but without the board and pieces, or learning music composition without sound. Statistics learnt solely from the pages of a book (or lectures) suffers the same problem, and un fortunately most students and researchers are expected to gain a practical grasp of the subject in exactly this way. A further problem is that much statistics teaching is based on mathematics, which is beyond the easy reach of most students (equivalent t o learning chess without a board in a foreign language). In addition, most statistics textbooks and courses are apt to utilise a few sets of data and work through them. This might sound acceptable, but students learn a handful of analyses, examples of s tatistics, rather than statistics itself.

Most of what we thoroughly know has been learnt by observation, trial and error. Statistics cannot be taught this way within the current mathematical framework using a selection of data sets, since the time required (and the number of prepared analyses) would be far too great. However, most students of statistics are not interested in it as an academic discipline. If the subject is instead approached as a tool to be used, a higher degree of flexibility can be achieved. Mathematical proofs become irrel evant; when learning to ride a bicycle, a child needs no knowledge of angular velocity, frictional forces or gravitational pull; an intuitive understanding of all of them will be impressed on him/her more or less painfully! Mathematical proof does not nec essarily lead to enlightenment, and much of what we truly understand requires no “proof” at all but repeated, varied and directed observation. This can be achieved in statistics teaching by making it into a game, after which it can be at least as interest ing as chess, and certainly a lot more useful.

Overview

Our Solution

 With this in mind we have approached the problem of teaching statistics by creating and computerising a series of challenges and games which the user plays by changing the data. This is a radical departure in statistics teaching, where typical statistica l practice is mirrored closely in making the data sacred and immutable. It is still possible to impress on students that in real life data cannot be changed; and by giving them the opportunity to experiment in the classroom in this way, they become exper ienced in recognising patterns and exploring strategies without danger, in a way which might take a decade or more to accumulate through real research. Traditional teaching can give students some variety, however the interactivity of our approach involve s them directly, making it more successful.

Should the results of the interactivity remain complicated, the student will simply have a deeper understanding of their confusion. By transposing the data into a simple graphical representation, whether a fitted line or a set of normal plots, the result s of interactivity become clear and the student gains a genuine understanding.

The data points are the chess pieces which may be moved in any direction. The student can be guided through a number of scenarios within which they are encouraged to experiment and observe changes in the resulting test statistics, finally being challenge d to generate particular outcomes. Although they may not understand the processes at first, by the end a clear intuitive comprehension will be established.

Overview

Conclusion

 We have been using Statistics for the Terrified software now for some time in the teaching of statistics, and those who have used it report that they feel much more confident in their ability to cope with the subject.

These results were extremely encouraging, not least because all users enjoyed the experience of learning statistics for the first time. The most effective learning took place when the data game was also accompanied by graphical representation, such as th e fitted line in the regression module or multiple normal curves (or Box Plots) in the Analysis of Variance module, although quizzes were also popular. The overall effect of the software was to produce an information funnel, with large amounts of raw dat a at one end, a small number of graphical representations and numeric statistics in the middle, leading to the final simple interpretation. This reflects one function of statistics itself as an informative summary tool.

Although the software does cover some quite advanced topics, it approaches everything in a basic, common sense way. Having gained an understanding at this level, the user is able to approach lectures and textbooks in a more informed manner. We do not fee l that our approach replaces the existing teaching of statistics, but it is a valuable precursor and accompaniment to it.

Areas covered by the software

 Simple data description (Mean, median, standard deviation, standard error etc.); Classification tables (Chi square); Repeated measurement (Paired t-test, area under a curve); Regression; Group differences (Analysis of Variance, two sample t-test, Mann-Wh itney etc) and Analysis of Covariance.

System Requirements

 To use the package you will require a PC which can run Windows, preferably with 8MB RAM and 256 colours.

Availability

 Version one is now available from the ITTI, price £25 plus p+p from the following address:
Mrs Jean Burgan
UCoSDA
Ingram House
65 Wilkinson Street
University of Sheffield
Sheffield S10 2GJ
Tel: 0114 272 5248 Fax: 0114 272 8705
E-mail: j.burgan@sheffield.ac.uk
Version two, incorporating additional subject matter as well as modifications from practical experience, will be available shortly from Radcliffe Medical Press, Oxford.

Steven Morris and Jill Szuscikiwicz
St. Georges Hospital Medical School
London
Email:-
S.Morris@sghms.ac.uk

[PREVIOUS] [NEXT] [CONTENTS] [CTI-HOME]