Issues with Maps
Base Rate Bias
Human-related events are correlated with population density.
For example, the map above may make the viewer assume that high density regions are more literate/culturally advanced. However, this is not true! They are simply more populated!
Insensitivity to Sample Size
When we have a large sample size, values tend to be stable around the mean. However, when we have a small sample size, values tend to oscillate and be either very high or very low. Smaller sample are, therefore, very unstable, when compared to larger samples.
Skewed Spatial Distributions
We can have very small regions and very large regions at the same time. This problem is particularly common while using the world map for a visualization: for example, Europe has very small countries while other regions have larger countries.
This is a problem because size has an impact on perception. Smaller areas are more difficult to visualize.
One solution to this problem is to use a Dorling Cartogram.
Perceptual Issues
It is difficult to estimate/compare values using color intensity
Color perception is affected by area size
Background color affects the perception of the color of foreground objects
The perception of the size of an object is also affected by the sizes of the objects that surround it
Map Projections
Every map is a result of projecting points on the surface of a sphere (the Earth) onto locations on a plane (the 1D map). This results in certain unavoidable distortions.
The following are some commonly used projection types:
We must keep in mind that:
larger areas are more prone to distortion
distortion increases as we move away from the point of contact (ex. for the cylindrical projection shown above, distortion is least at the equator and highest at the poles)
Some rules of thumb:
for equatorial regions, use the cylindrical projection
for mid-latitude regions, use the conic projection
for the polar regions, use the azimuthal projection
Tissot's Indicatrix is a measure of distortion. The idea behind it is to start by placing circles on the original surface and then measure how the circles get distorted when they are transformed to a plane.
The Mercator Projection is the most widely used projection. It has no angular distortion.
However, there is area distortion from North to South. The areas in the North and South are much larger than the areas in between.
The image below shows the Gall-Peters Projection. It tries to preserve area but there is visible shape distortion.
Finally, we have the Robinson Projection. It doesn't completely preserve any specific feature, but is a good compromise among all the features.
Equivalent: preserve equivalent areas, useful for world maps and small scale in general
Conformal: preserve angular relationships, useful for large-scale maps
Interference from Map Features
Sometimes, the features of the map in the background (such as region boundaries) may interfere with the actual objects that we want to visualize.
We must make sure to keep the focus on the spatial distribution of the values contained in the data. To do so, it is important to decide how many and which details from the map we want to retain.
If we use too many map features, they will interfere with our data. However, if we use too few map features, we may lose important contextual information. There is always a trade off.
A few rules of thumb:
try to use a few colors (preferably with low saturation); greys are great; don't use too many colors
do not use string lines or borders
include only necessary spatial features
Last updated
