Visual Data Exploration: Tools, Principles, Problems

Our 1999 IJGIS paper (Andrienko and Andrienko 1999) was about maps as tools for exploratory analysis of spatially referenced data. After the paper was published we continued working on tools and techniques for exploratory analysis of spatial and, more recently, spatio-temporal data. We have written many other papers concerning this topic and have just finished writing a book, in which we try to approach the topic in a systematic way and to lay theoretical and methodological foundations for exploratory data analysis. Exploratory data analysis (EDA) means open-minded looking at data with the aim to detect and describe patterns, trends, and relationships and to generate hypotheses, which can then be tested using mathematical methods. The very nature of EDA implies that data visualisation plays a crucial role. As is defined in a dictionary, " visualise " denotes " to make perceptible to the mind or imagination ". In EDA, it is the mind of a human explorer that is the primary tool of analysis. It is the task of the mind to detect and describe patterns and to generate hypotheses. However, the mind can fulfil its mission only if the data to be explored are made perceptible to it. No thinking is possible without prior perception. Hence, data visualisation is the most important supporting tool (i.e. supplementary to the human mind) in EDA. For the perception and further reasoning to be valid, a visual representation of data must be structurally similar (isomorphic) to the pertinent features of the underlying phenomenon (Arnheim 1997). A specific implication of this requirement is that geographically related data can be properly explored only with the use of maps as isomorphic representations of the geographical space. It was therefore natural that we started our research into tools for exploratory analysis of spatial data with maps as primary means of visualisation of such data. In this way, our research interests became very close to those of the Commission on Visualisation of the International Cartographic Association (see http://kartoweb.itc.nl/icavis/). In fact, what we did was fully in line with the research agenda of the Commission, with its " emphasis on the role of highly interactive maps <…> at hypothesis generation, data analysis, and decision-support " (MacEachren and Kraak 1997). Therefore, the community of researchers in geovisualisation clustering around the Commission enthusiastically accepted our work and adopted us as its integral part. It is interesting that this community does not consist …