Stat. Graphics & Data Visualization

Although the topic "User Interfaces" appears in the title of the blog, I didn't post anything related yet.

Moving into a new office building confronted everybody with the new switches in the offices which control light, temperatures ...

Here is what they look like:

(Sorry for the fuzzy photo, but the 2MP cam in my SE-phone isn't capable of more ...)

The only thing that works as expected here are the light switches, which toggle between ON and OFF - except that you always have to read the text in order to find the desired kind of illumination.

Touching the buttons of the shades makes them go up or down as expected, but the problem arises when you try to stop them. Hitting the button again does not do the job, hitting the button for the other direction reverses the direction but does not stop them either - hmmm?!
The solution: Pressing the buttons for more than 1 second will stop them when you release the button. Unfortunately the delay of 1 second is about the average time people will press the button, resulting in an apparently random behaviour.
The porthole-like windows open and close when pressing the corresponding button - this time they only move as long as you actually press the button.

The highlight is the temperature control. You can press 'cooler' or 'warmer' and ... wait. One very important issue in user interface design is feedback. I actually could not figure out how the temperature control works until I read the manual!
The feedback works as follows:
There will be a red light next to 'cooler' when it is the standard temperature of 21C. When you press 'warner' once it will be 1 degree warmer indicated by a red light next to 'Window open', and once again, i.e., 2 degrees warmer light the LED next to 'Shades up' (as in the picture).
The whole thing works for cooler as well, but now the LEDs will flash to indicate that it is cooler than standard ... The only purpose of the green LED is to indicate that the AC is actually working, which won't be the case as soon as you open the window.
Would you have guessed?

So much for intuitive controls ...

This time it is easy for me, as I just need to point to the interesting discussion at Junk Charts.

The original graphics is from the NYT:

The "improved" plot looks like this:

The advantages of the box-plot view (from Junkcharts)

• The European market is much more fragmented than the U.S. market.
• The Big 2 (GM, Ford) has had mixed fortunes over this period (as indicated by the large variance)
• The Big 2 are competitive in Europe although they are definitely not dominant there
• Several key players in Europe (Peugeot, Renault, Fiat, BMW) have negligible shares in the U.S

The discussion was quite lengthy, but had the two major points:

• Boxplots are too hard to read for NYT readers
• The boxplots ignore the temporal information and thus are not really suitable for this data

One important point was not mentioned explicitly, which is

• Sorting is a very powerful and important option in graphics

There is truth to all the issues raised here, and the bottom line is probably that there might be not a single graphical view on the data which covers all aspects of the data. Furthermore, the NYT graphics is by far the most eye-pleasing version of all suggestions.

In this spirit I want to throw in two more versions which show the data:
Two variations of Mosaic Plots (will be explained in the next 'statistical graphics 101') in the multiple barchart view. European cars are highlighted:

(Year x Brand x Continent)

(Continent x Brand x Year)