|Title||Statistical Analysis: A 'Project Approach'|
|Department||School of Geography, Kingston University, Penrhyn Road, Kingston upon Thames, Surrey. KT1 2EE|
|Tel.||+44 (0)181 547 7512|
|Fax.||+44 (0)181 547 7497|
Students of all ability levels, but especially those with a non-scientific public examination background, often have difficulty in appreciating the relevance of statistical analysis to geography. The 'project approach' to teaching data analysis and statistics gives students a sense of ownership over the data they analyse and relates learning about statistical methods to their systematic and thematic geography modules. The aim of the 'project approach' is to develop an understanding of statistical analysis that places it as one of a series of activities or tasks that contribute towards the development of geographical knowledge and explanation.
The practical project, which uses computer software for statistical calculations, lasts for 9 hours over three weeks depending on timetabling and student numbers. Each supervised session would relate to a separate stage of the project:
This approach to teaching quantitative methods would be amenable to various human and physical geography projects, and could be integrated with fieldwork. It has proved successful with a project where the aim is to estimate the areas of different categories of land use across a 20 by 10 km map from the Second Land Utilisation Survey (note that the full range of land uses are collapsed into 10 principal types). These maps, which were produced following field survey work in the 1960s, exist for most parts of England and Wales. Setting the sample size at 20 km squares, the project design stage can explore a range of different sampling strategies based on eastings and northings being selected with variable degrees of randomness. Possible strategies include 20 random eastings and northings within the ranges on the map, two 10 km vertical transects determined by randomly chosen eastings, one 20 km horizontal transect using a northing and random selection of two eastings per northing. In each case a transparent overlay marked with a 1 km square divided up into a 10 by 10 ha grid is used to estimate the areas of each land use per square. The data generated are entered into suitable statistical software such as MINITAB or SPSSX.
The second stage involves the generation of various descriptive statistical measures (means, ranges, variances, standard deviations, etc.) and the production of absolute and relative frequency distributions for the 10 land use types. On the basis of the known sampling fraction (10 per cent) it is also possible to estimate the area of each land use across the whole map sheet by simple calculation. The final stage entails presenting the students with some additional information from the published statistics derived from the Second Land Utilisation Survey relating to the county in which the project map is located. These constitute population parameters against which the students' estimates can be tested as an introduction to hypothesis testing, using procedures such as Z or t tests. The project concludes with a group discussion in which differences in the estimated area of the 10 land uses as derived from the various sampling strategies forms the focus of attention.