Small multiples re-imagineered

This chart gave me trouble.  I kept staring at it, staring.  Searching for the legend.  What could the several lines, in different colors, represent?  Take a look yourself.

Well, it turns out all three graphs were duplicates.  A different line was given dark blue to highlight a particular amusement park.

I have not seen this tactic used before.  This is like a small multiples concept except that every chart contains the same data.  Is it better than having just one chart?

Reference: "Will Disney Keep Us Amused?", New York Times, Feb 10 2008.

PS. [4/6/2008]  Here are two alternative charts contributed by our readers.  See comments below.

Derek suggested using sparklines:

Zuil reverted to basics:

Comments

[Clint]

I was just confused...
Don't make me think!

[mike anderson]

I'm not sure whether the technique is Goofy or Dopey. It does suggest overlaying the three charts--say in an animated GIF--for a computer display, which would be interesting, but for print, nah.

[Nick Barrowman]

I think it's a terrible waste of space, and not at all intuitive.

[derek]

I noticed the other day that The Economist had apparently discovered dot plots (see the "click to enlarge" panel), so it seems that large publications are paying attention to calls for more graphical sophistication. But they sometimes get too clever, and here I think the NYT did: this technique is a good one for untangling e.g. nine complex lines, but for three simple ones? The ordinary colored lines approach would have been fine.

See the comments in this Juice Analytics article for example. This is about handling hundreds of lines, where different colors would have been a disaster. I look forward to the New York Times actually presenting a data set complex enough to require the approach they used for this graph type, instead of such a simple set.

[Harlan]

I agree with Mike. This would be useful for a Powerpoint presentation (or for when the Times is printed on Smart Paper and can use animation!), but it's just confusing now...

[Jon Peltier]

They tried to use a static display to simulate a dynamic approach, where the highlighting is based on which item in a list is selected. The dynamic approach works online, where a reader can interact with the display. In print, it was confusing.

[timThompson]

Derek, the Economist has used dot plots to represent multi-level opinions (in this case by country, in others--such as the annual Latino attitudes to democracy article-- by year.)

So I'm afraid this doesn't represent any progress.

I think dot plots may be the presentation format of choice of the Pew polling company the Economist uses, since they only seem to occur when the magazine reports opinion poll results.

[Joe]

I agree with the general consensus here, it's an acceptable technique for complex data sets, but in this case it's a waste of space.

[Siddharta]

Useless. Small multiples are meant to show variation along an axis (thus useful for multi variate data). But here all three graphs are exactly the same. Whats the point?

Plus, like everyone else, I didnt notice that they were same to start with and got confused, so its actually _worse_ than one chart.

[ZBicyclist]

If we all just agree, we won't have discussion. So I'll defend the graph.

Note each graph is part of a column of facts on a particular park, including year opened, number of rides, location. This is the "time trend of attendance" element to that table.

Showing the time trend of attendance as a graph is good. Showing the other two lines faintly shows (1) an easy comparison across parks, and (2) that the same y axis is used for all.

I think that's the logic of using this technique: the graph is part of the table. But due to the overall design elements of the table, the fact that it IS a table gets lost.

[derek]

They're sparklines that forgot that sparklines are supposed to be the size of a word, not the size of a graph?

Also, there's a common error of table design here. Having the table fields be in rows, and the records be in columns, means that in each column a long label, e.g. "Lake Buena Vista, Fla.", is inefficiently sharing a column with a short four digit year, e.g. "1971".

If the records had been in rows and the fields in columns, the table would have been properly orientated according to our European left-to-right writing system (the opposite rule goes for East Asian top-to-bottom writing systems).

Then the sparklines could have been in a column of three rows, eliminating the need for gray comparison lines.

[derek]

Example

[zuil serip]

I am not sure that sparklines are a good solution here.

A sparkline is a space efficient way of displaying a time series for a single variable for a well defined period of time.

In this case, however, we also want to see the relationship across the three time series. Ie. Magic Kingdom and Disneyland have roughly the same level of attendance, while California Adventure is roughly 1/3 of that.

The sparkline completely obscures that - for example, it is impossible to tell whether MK or Disneyland had higher attendance in the beginning of the period. Even the beginning and end dates for the period are undefined.

I think the original chart can be fixed with much less radical surgery. Perhaps, something like this:
http://www.flickr.com/photos/24579696@N05/2379879167/

[Doug]

derek and Jon Peltier are right: They're taking a dynamic chart and showing static frames, like displaying cells from an animation, which may be natural enough to Disney employees, but not me.

I like zuil serip's version.

[Jon Peltier]

I prefer Zuil Serip's annotated chart over the sparklines, for the reasons he posted.

[derek]

It wasn't actually my intention to approve sparklines as a best choice in this case, only to suggest those sparklines as better than the table in the original, as glossed by ZBicyclist.

Does that make sense? I mean it in the same spirit that someone might suggest bullet graphs as a superior implemention of a single-number graphic than speedometer graphs, while reserving the right to disapprove of the original decision to construct a single-number graphic :-)

[Kaiser]

Derek: I'm with you. That's the spirit we like to have on this blog. A lot of charting is experimental; without seeing different versions, it's hard to judge which does the best job. Keep sending us new charts!

[Jon Peltier]

I wasn't criticizing the sparklines, as Kaiser said, it's all experimental. The sparklines may have benefits over the original graphic, while the chart by Zuil Serip has others. Usually there's no single chart that shows all the interesting things about a set of data.