Posts tagged with graphics

Everything in computers is so much more complicated than you think.

(Source: Wikipedia)

hi-res

These charts are undeniably beautiful, but they violate Tufte principles 1, 4, 7, 10, 11, 12.

Charts can look great but E Tufte says we should let the data do the talking, rather than the design. Adding some sparkle to the data is “wrong” or at least, Tufte-wrong, for data-graphics.

Here it seems like the talented artist has tried to “add some sparkle and theme” to “boring numbers” — rather than accentuating what’s exciting about the numbers themselves. To my way of thinking, if the message the numbers are telling you is interesting, then that makes the numbers worth looking at.

• Did you say I could get a 25% raise?!
• Did you say people are 30% taller than they were 250 years ago?
• Did you say a 19% chance of rain on our wedding today? Or 90%?
• Did you say the cost of electricity is one-one-hundredth of what it was 90 years ago?
• Did you say my heating bill is double what it needs to be if I insulated better?
• A man and a mouse are only one order of magnitude apart?
• I could commute across America on a bike if I were two orders of magnitude faster?
• Did you say that 99% of the people own 1% of the wealth? Or was it 99.999% of the people owning .000001% of the wealth? Or both? Wait, these numbers are actually crucial to the story!

Of course it’s no surprise that most people think cifras son aburridas — since their main memory of figures is through boring maths class, rather than as integral elements of a story.

As in the wieners I drew, it’s not easy to make the logically beautiful look visually beautiful.

My interpretation of [Leland Wilkinson’s] grammar [of statistical graphics]:

Data is the most important thing, and the thing that you bring to the table.

—Geometric objects … what you actually see on the plot: points, lines, polygons, etc.

Statistics transform the data in many useful ways. For example, binning and counting to create a histogram….

—Scales map values in the data space to values in an aesthetic space, whether it be colour, or size, or shape. Scales also provide an inverse mapping: a legend.

—A coordinate system describes how data coordinates are mapped to the plane of the graphic. It also provides axes and gridlines to make it possible to read the graph.

— A facetting, or conditioning, specification…. to reproduce the same plot for different subsets of the data. The facetting specification describes those subsets and how the facets should be arranged in to a plot.

`cumsum ( rnorm(50), lend="butt", lwd=12, type="h" )`

Cumulative sum of 50 draws from a normal distribution.

File this under mysteries of the Central Limit Theorem.

hi-res

Here is how to improve your charts, graphs, maps, and plots:

• Erase non-data ink.
• Erase redundant data ink.
• Maximize the ratio of data to ink.
• Show data variation, not design variation.
• The surface area of graphical elements should be directly proportional to the numerical quantities represented.  (Don’t use 3-D bar charts, for example.)
• Don’t lie.
• Get as much data as you can in the first place.
• Apply the right transformations to the data (adjust for inflation, divide to per-capita numbers, take the square root of naturally squared quantities).
• Then, you can shrink the graphics way down.
• Increase data density and data resolution.
• Maximize the amount of information per unit of space.
• Maximize the amount of information per unit of ink.
• Above all else show the data.

For example, here’s how he would use the eraser, not the pen to improve on the typical bar chart or histogram.  (3-D bar charts are right out.)

Additionally, Tufte wants news publications to use sophisticated graphics that let the data tell their intricate story, rather than simplistic graphics that attempt to “dazzle” the viewer.

• Like good writing, good graphical displays of data communicate ideas with clarity, precision, and efficiency.
• Like poor writing, bad graphical displays distort or obscure the data, make it harder to understand or compare, or otherwise thwart the communicative effect which the graph should convey.

Lastly, regarding wide versus tall graphics:

• If the data suggest a shape to the chart, follow that suggestion.
• Otherwise, move toward graphics about 50 percent wider than tall.

hi-res