Map type
The Gastner-Newman Cartogram (otherwise known
as a population-density equalising cartogram) is a technique for
representing data for areas that modifies the size of the area to take account
of different denominator populations. It’s an approach that allows size of the
area to be equalized so they are visually comparable. The most common approach
for this contiguous area-based cartogram is that developed by Michael Gastner
and Andrew Newman which does a relatively good job of preserving shapes even
though areas are quite distorted.
A cartogram can be a powerful approach to mapping population data since it provides a strong visual for numerical area data and does not require data to be normalised. The drawback is that the distortions can be a hurdle for some in their understanding. Data classification and symbolization remain crucial and many of the principles that apply to a choropleth map apply equally to designing a good schema. In this example of the 2012 Presidential election, the map is designed to show the percentage share of the vote gained by the Democrat and Republican partiesat county level by using a blended hue technique.
Data
The Gastner-Newman technique can be used to show both absolute values or data that has been normalised since the problem of unequal areas is accounted for in the technique itself. A single value should represent each area.
Data will normally be classified using a suitable scheme. Natural breaks is a good default as it tries to group similar data values into the same class as a basis for defining class breaks. Alternatives such as equal interval and quantile can be useful in some circumstances. Here, the data are classified into 20 classes using an equal interval scheme that gives percentage categories each of a 5% class interval above and below a 50:50 share of the vote. Ordinarily we wouldn’t have more than 5 or 6 classes since it becomes difficult to differentiate between many more classes on a map. However, this map is being published as a web map so the ability to click the counties and retrieve the actual data value allows us to stretch our symbol palette to see more nuance in the data.
Symbols
Symbols should be designed so that different magnitudes of data, represented by unique classes, can be easily distinguished from one another through variation in the symbol’s hue, lightness and saturation used as ordering visual variables. For this example a blended scheme is used that varies from pure red for one end of the Republican classification, through purple for areas that have a shared vote to pure blue for the Democrat end of the scheme. Using blended colours like this can be useful when the colours themselves signify meaning in the data.
Unless a specific colour suits the theme being mapped then neutral hues should be chosen to avoid making a map with value-laden symbology. it is also important when using a diverging scheme like this to ensure each end of the scheme is no more visually important than the other else it will dominate the map.
Marginalia
The legend should include a representation of the symbol classes or a frequency distribution graph to enable readers to understand the classification and symbolization scheme. Here, the legend is embedded in a popup and describes the relationship of the symbology with the frequency of values for each map. The map should be finished with a succinct title, source details and relevant credits. Popups are used to allow the map reader to recover the actual values for each county which overcomes two of the problems of estimating class values on a static cartogram: namely correctly identifying the class that each area falls in; and the inability to recover individual values.
Map Use
When viewed, the map reader should be able to efficiently recognise the different
classes mapped across the map and to see the areas of high and low, while at
the same time seeing which end of the spectrum they are viewing. At the very least, relative differences
should be obvious and the reader should be able to determine a pattern across
the map. We visually interpret the
symbols as tending towards the red or blue ends of the spectrum so we perceive purer
colours as a greater share of the vote for that candidate. The multiscale
characteristic of a web map means we can design the map to reveal more context
by adding labels at larger scales to aid interpretation.
