Also available as an Acrobat File |
Software Evaluations |
The Visualisation of Area-based Spatial Data3.1.2 Graphical effectiveness of toolsThe second key area of functionality is the effectiveness of the graphical facilities in assisting and encouraging the exploration of the data and activities such as hypothesis formulation and testing. In order to develop a set of criteria for making this assessment we draw upon ideas from the literature on statistical data graphing and cartography. Surprisingly little attention seems to have been paid to the assessment of the effectiveness of different graphical techniques used in modern visualization software. Cartographers, who have a long history of assessing the effectiveness of different designs for traditional printed maps, have only just begun to address this issue (Monmonier 1991, MacEachren and Kraak 1997) with the establishment of an ICA commission on visualization (ICA 1997). This means that we have had to define a set of guidelines ourselves, building on existing work on statistical graphing (Cleveland 1994), the design of graphical material in general (Tufte 1983, 1989) and the design of paper maps (Buttenfield and Mark 1994). Cleveland (1994) provides a useful classification of the tasks required by an observer to interpret a statistical graph, which will be extended for the purpose of this case study to the interpretation of maps. According to Cleveland there are two main ways in which a graph displaying either quantitative or categorical data may be read:
The purpose of the classification is to provide a basis for evaluating how well any particular graph performs in portraying data. The six activities can also be related to many of the principles of good design which have been identified by Tufte (1983,1990) for graphics in general and Dent (1985) and Cuff and Mattson (1982) for maps. If there are differences between pattern perception on graphs and on maps they may lie with respect to the types of pattern properties that are looked for on maps and the extent to which the individual (areal) units differ in size and hence contribution to detecting overall pattern. In a series of influential books, Tufte (1983, 1990) has developed a series of principles which should guide the development of good graphics. The key one, underlying all the others, is that a good design should show the maximum amount of data using the least amount of ink in the smallest space. An effective graph or map should focus the viewer's attention on the data being displayed, and not on the attractiveness or otherwise of the design itself. By aiming for a compact presentation, it is possible to put more data on a single page, or the same amount of information in a smaller space, both of which facilitate comparison between data values. Essentially the points Tufte makes about good design all serve to assist the reader in undertaking the six tasks defined by Cleveland (1994) - graphs should have clear annotation, with consistent and logical symbolism enabling the reader to make unambiguous interpretations of the values and patterns in the data. Similar comments may be made about many of the general rules of map design. These are covered at length in standard texts such as Dent (1985) and Cuff and Mattson (1992) but Buttenfield and Mark (1994) provide a very useful summary. They define the main stages of map production, and the rules governing each as follows. Generalisation - this includes the selection of an appropriate amount of material to display so that the final map is not visually cluttered. With maps, it is also generally accepted that some distortion of reality is often needed - for example roads are often simplified, and buildings moved from their true geographical location so that they lie alongside the simplified road course. It is therefore common in maps to `lie a little to tell a better truth' (Monmonier 1991) something which appears to be less common in drawing statistical graphs but occurs for example when "jittering" a graph to reveal overlapping cases. Brassel and Weibel (1987) also identify statistical generalisation - the simplification of data to portray trends rather than details - as part of this process. Symbolisation - this includes the selection of appropriate symbols (point symbols, line types) and colours to portray the various map elements. A series of `rules' have developed about appropriate symbolism, some based on features of the human perception system (e.g. red is perceived as a `stronger' colour than blue), some on natural associations (e.g. blue for water), some on convention (e.g. brown for contours on UK OS maps) and some on principles of design (e.g. the use of colour `sequences' to portray sequence in numerical data). Production - the final stage includes organising the geographic and non-geographic map elements (title, key, scale) on the page to achieve a balanced overall design in which the main features of the map (i.e. the data in thematic maps) is visually most prominent. It is clear that the design rules for statistical graphs and maps have many features in common:
| |
Presentation | Visualization | |
---|---|---|
Static | Dynamic | |
One view of data | Multiple views of data | |
Data unknown to viewer | Data known to viewer | |
Table 3: Comparison between features of presentation graphics and visualization. However, as noted above, all these rules have been developed in the traditional world of presentation graphics, which has a number of different characteristics from the world of interactive graphic visualization software (Table 3). This means that some of the rules developed for paper maps may no longer be appropriate. As a simple example, maps for visualization do not need a north arrow, scale bar or title (normally considered essential for presentation maps) because the analyst can be assumed to be familiar with the study area. In general the move to interactive graphical presentation will produce two types of change:
It is clear that a good deal of work remains to be done to examine just what rules are appropriate for graphics in visualization packages, and to assess the effectiveness of the graphical tools which exist at the present. Such work will become even more urgently required if one outcome of the current AGOCG project is to try and encourage a greater use of such tools among social scientists in general. |
Graphics Multimedia Virtual Environments Visualisation Contents