Mapping multivariate statistical data is fraught with difficulties due to the problem of encoding multiple pieces of information into a coherent yet simple symbol design. The balance between making your data readable and understandable is harder, the more pieces of data you want to show. It’s also very easy to end up with symbol overload that easily translates to cognitive overload.
In 1973, Herman Chernoff an applied mathematician, published a paper in which he proposed the use of a human face to display multivariate data in the shape of the human face. The idea was simple – change the shape, size, placement and orientation of different facial features to encode different variables. The overall result changed the face and the facial expression which brought an overall sense to the combination of data. Chernoff argued that small changes can be seen in human faces due to our innate ability to recognise human faces and their subtle differences. They are extremely economical graphical structures in which
Of course, choosing how to reflect the data in different facial features is fundamental. Here. Eugene Turner creates what has become a well used example of the genre, though the use of Chernoff Faces has courted controversy and not seen a huge uptake largely due to difficulties in construction. The additional problem is in the way in which we automatically interpret the faces through our impression of emotion. Hence the crucial need to ensure your variables are mapped in a way that works rather than one which conveys the wrong emotional response. Turner does a good job of building a facial profile out of social conditions and ethnicity. It’s a simple map but one that characterises the spatial structure of socio-economic life in Los Angeles. It’s also a provocative and arresting image and one which is difficult to hide from.
Chernoff faces deserve a mention in any design related commentary because they are innovative. They’re hard to employ correctly but even if only a small proportion of our data can be effectively mapped using them, they’re still a useful tool in the cartographer’s design armoury.
Chernoff’s original paper can be downloaded here.
Although strictly speaking MapCarte is about showcasing great examples of map design, many maps from after the mid-late twentieth century and onwards owe much to one man’s detailed exposition of cartographic symbology. Jacques Bertin’s ‘Semiology of Graphics‘ was first published in French in 1967 and subsequently translated into English and republished. Bertin’s classic ‘Semiology of Graphics’. It is a classic of information design as it leaves almost no stone unturned in presenting the principles of graphical communication.
Bertin developed many standard rules which have become de facto syntax and grammar of the cartographic language. He didn’t make many maps outside of his book but the second part contains over 1,000 maps and illustrations that demonstrate the clarity of the principles at work Here, we see a set of maps of France showing how one can manipulate some basic graphical marks to present multidimensional (multivariate) data. This, of course, is one of the challenges for any thematic map and usually the answer is to eliminate detail and present different variables using separate maps. However, with careful choice of symbology it’s possible to combine symbols to represent multiple pieces of information in a way that doesn’t clutter the image or cause the map reader to be overwhelmed.
A classic text with many perfectly framed maps and illustrations that showcase the graphical language of cartography.
Click the image to view the online web map
Following the mapping of elections has become almost as much of a spectacle as the elections themselves. Media outlets seem to fall over themselves trying to provide more detailed and expansive coverage than their rivals. Over the course of the last decade we’ve seen countless ways to map the detailed voting patterns as they are announced and the map takes shape. Proportional symbols, choropleths, a variety of cartograms and three-dimensional techniques have all been used in recent years.The map presented here hasn’t been widely used and takes a different graphical approach altogether.
Made as a web map and published by the online arm of German newspaper Die Zeit (The Times), the Wahlland map shows the Germand Bundestag elections of 2013. The electoral districts of Germany are arranged by their voting behaviour using a very abstract approach that is based on similarity of lifestyle, preference and attitude rather than geography. Each district is represented by a series of triangles arranged almost like a coxcomb chart over each district. Each segment maps the results (by size) for the party represented (by colour). Rather than simply mapping voter numbers the size maps the relative difference to the average result of that party overall. Only parties that are above average are mapped and the districts are arranged so that more similar districts are clustered. It’s a novel approach to mapping multivariate statistical data that shows not only the salient information of the results but in a way that adds value and provides some sense of the underlying geography and patterns being revealed. Which districts voted like each other? Are they geographically proximal or distant?
The design goes beyond by providing a very helpful legend (hidden until required) and also allows data to be seen through hovering over each segment. The map can be filtered by state or by party so the clusters are easily viewed and the subtle background shading works to anchor the eye to a cluster and give a hint as to the overall quantity of vote for a party. Geography is thrown to the wind but this is a novel and well crafted alternative to more traditional forms of mapping elections that is visually eye-catching and performs well. Even the way it draws is neat, smooth and easy on the eye.
A static thematic map showing active social networkers. More often than not, maps take complex data and omit a great deal of information to present a single message clearly. Sometimes, however, clarity is achieved through a deliberate design that integrates a number of variables so that the multivariate character of various inter-relationships is preserved. These maps are designed to be studied beyond a simple glance. They contain detail and do not apologise for demanding that the map reader explore them. Often, the design challenge is to communicate complexity simply and by mapping multiple variables you inevitably make the job harder!
Here, Rikard Andresen uses a range of graphic devices and successfully creates a simple, clean and effective map. Bar charts show quantities and Venn diagrams show how they overlap and interact. The use of simple colours is a logical symbolisation that feeds off people’s innate understanding of colour. The Venn diagrams are proportional and actually function as small multiples allowing easy visual comparison with each other.
The map itself is almost immaterial to the overall display but as a subtle background it helps to lead the eye to different countries. By necessity, symbols are not placed directly across countries but that doesn’t matter. What Andresen creeated in fact, is a discontinuous Dorling cartogram, presented with a visual link to approximate location to aid our ability to peruse logically. The title is unobtrusive, the surrounding detail supports the map. The main map is allowed to take central stage.