<-- Advertise here.

a novel data visualization method for visualising the changes over time in the ranked order of any set of objects where the ordering is usually from large to small.
this technique has been used to illustrate the size of cities in the US from ad 1790, the UK from ad 1901 & the world from 430 bc. the resulting visualizations destroy any notion that rankÔÇôsize scaling is universal: at the micro-level, these clocks show cities & civilizations rising & falling in size at many times & on many scales.

[link: nature.com & ucl.ac.uk]




How do you read these things? It's not immediately apparent to me.

Wed 06 Dec 2006 at 8:12 AM

I share the previous commenter's concern. It isn't obvious how this graph should be interpreted. For instance, what do the colors represent? Part of the problem is that the images are too small to closely examine.

A more fundamental question regarding this diplay is, why would you want to display linear time in a circular fashion? Wouldn't a normal line graph communicate this message more effectively?

Wed 06 Dec 2006 at 9:50 AM
Stephen Few

A circular time scale makes for a nicer-looking and more compact display - but it also suggests something that's periodic, even if the actual data is not. I guess people also really like the clockface metaphor for time.

It's unfortunate that aesthetics are often more important that the finer points of meaning, especially when it comes to communicating to an audience that does not know a whole lot about visualization. This is giving aesthetics a bad name among visualization people, when it could be used for many useful things, too.

Wed 06 Dec 2006 at 11:25 AM

I also agree with all previous comments.
To me the visualization suggests that the difference between 1990 and 1790 is the same as between 1790 and 1800.

Wed 06 Dec 2006 at 3:47 PM

for those wishing to examine full-scale figures, the Supplementary Notes of the Nature paper is available here (pdf).

Wed 06 Dec 2006 at 4:54 PM

to robert: I am not so sure whether the original authors had specifically any visual aesthetics in mind, except of maybe the rainbow color scheme.

Wed 06 Dec 2006 at 9:02 PM

They may not have had aesthetics in mind explicity, but I think that the looks of the "clock face" made it a more appealing choice, as compared to a simple line chart, for example. Line charts just aren't sexy, so doing something more interesting is probably necessary to get a paper into Nature (and many other publications).

Thu 07 Dec 2006 at 1:24 AM

Rank clocks must be read in relation to rank size, that is the whole point: In other words - the order implicit in the scaling laws as measured by ranking the objects in terms of size is completely obscured until one extracts their temporal volatility. In the Nature paper, there is an example of how confusing the picture is if you simply plot changes in rank over time in rank size space. Plotting these in clock fashion is more economical than in linear time although as one is a simple transformation of the other, then no more information is imparted. Clocks do impress the idea of cycles more easily than line graphs in my view and also indicate much more easily how objects shoot in and out of the rank spze with respect to time. In the article there are also a number of clocks which involve growth rates and their disaggregation which again in my view are better visualised as clocks.

The computer program that visualises these clocks also of course has visualisations in terms of linear graphs and many other transformation and the program can be unlocked to show the line graphs so that readers can judge for themselves.

Lastly the critical theoretical issue in this article is that one should examine the microodynamics as well as the macrodynamics. That again is the point for there is nothing sacred about the clock idea but there is real interest in the fact that such a regular macrodynamics obscures a volatile microdynamics.

Fri 08 Dec 2006 at 4:25 AM
the author
Commenting has been temporarily disabled.