A post from data.visualisation.free.fr
Above all, show the data
In his seminal book,
Edward
Tufte proposed some rules for good visualisation practices. He
encourage truthfulness, some sort of data transparency allowing the
reader to "see the data" and even defined a "lie factor" as the ratio
between the printed representation of the data and the real underlying
values in the dataset. Nobody doubt that one should design a data
visualisation with honesty, and avoid any temptation of cheating or
little arrangements with the data. But some did, and there were right
to do so!
Early ``cheaters''
Charles Minard’ (1869) Napoleon’s Russian campaign
map graph was quoted by E. Tufte (again) as “
the best
statistical graphic ever drawn”. This is probably true, and
Minard was a pioneer in thematic cartography and wanted to let the data
“
speak to the eyes.” Hence, we can spot some approximations in
the design of the map, its lack of projection reference (as quoted by
Michael
Friendly), but also some arrangements in the aggregation of flows
leaving and then rejoining the “great army”. This looks really nice and
is really a great map.
Some approximations for a storytelling map (a brilliant one)
In reality Napoleon’s army was divided into several branches which
followed different paths, but which are not showed on Minard’s maps
for clarity (see Martin Grandjean for details). So, Charles
Minard simplified the data representation on purpose.
Simplifying is a trick to see through darkness
Note that simplifying, or aggregating is an interesting approach for
visualizing complex data, such as flows or networks. A new and quite
promising technique, based "bundling
and shading" emerged recently and is aimed at simplifying the
visual representation by aggregating similar flows. An illustration is
given below with the work of Christophe Hurter on Airline's network in the
US. On the graph on the right, the visual representation has changed so
that possibly none of the airline used those aggregated trajectory. They
do not correspond to a real movements observed in the dataset but are
useful patterns to see through network data darkness.

Usual ``cheaters''
No map is perfect. Behind this innocent sentence is a very annoying
fact: It is, in fact impossible to truthfully represent a 3D surface
on a 2D plane. There must be some sort of distortion, and this is very
clear to anybody having to produce a map. But there are maps that do
not reflect the physical really of the underlying data source and
still very useful because of the simplification. Subway maps
Subway maps are oversimplifying the geography to highlight useful
information. And only that.
Benjamin
Schmidt has showed that New York has one of the most
geographically correct subway map, except for the cardinal directions.
But that's generally not true.
Another very popular and very popular cheating is done by artistic
view provided on ski resort maps . For a very long
time, ski maps were conceived and produced as drawings. Ski maps were
an artistic view of a quite complex reality. The art was complex, as
it needed to provide a map where people could locate themselves,
identify summits, slopes, and points of interest.
Cheating on ski resort maps is necessary for orientation
So these are maps, in the conventional terminology. But providing a
simplified information in a complex landscape implies to distort
reality in order to show both side of a mountain, uphill and downhill
perspectives, shadows, landscape, vegetation, etc.. Pierre Novat was a leader in this field in
France and abroad for many years. His work is based on a subtle mix of
maps, photographs, pencils and paintings. It has to be "realistic and
to provoke a ski desire". So he was cheating, as everybody's knows, of
course, and this was useful.
In my next post, we'll see why cheating is also necessary in data
science.
Done in Toulouse (France), by Xtophe.
Usual citation policy and disclaimer apply. Comments on my twitter account are welcome
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