by Edward R. Tufte

—

In january I attended his workshop, and I was really very impressed by the way he manages to capture your attention and keep it. The chance to observe him form close-up (we got a spot in the first row) was something we loved. On a critical note, they way he explored his format didn’t allow for a lot of discussion. Granted, there were far too many people to make a discussion very practical, but once the tutorial is over you have to get out there and make it on your own two feet. So how could I manage to impress on my managers that powerpoint can be a really inappropriate format sometimes?

One of the follow-up things for myself was initiating the discussion about how we communicate within Research. Most of the time, one has to make one-page statements and such. Those one-page statements then have to fit on a powerpoint slide… Not my favourite program for any such things. Also, a lot of stuff that goes top-down only comes in ppt, like they think we can only understand things when they come in bullet-points. Myeah, that’s we all at least have a Masters in something or other! There’s a time and a place for everything, and I wish that upper echelons would be more conscious about their communication methods.

—

Direct quotes from the book:

Graphical integrity is more likely to result if these 6 principles are followed:
1. The representation of numbers, as physically measured on the surface of the graphic itself, should be directly proportional to the numerical quantities represented.
2. Clear, detailed, and thorough labeling shuld be used to defeat graphical distortion and ambiguity. Write out explanations of teh data on the graphic itself. Label important events in the data.
3. Show data variation, not design variation.
4. In time-series displays of money, deflated and standardized units of monetary measurement are nearly always better than nominal units.
5. The number of information-carrying (variable) dimensions depicted should not exceed the number of dimensions in the data.
6. Graphics must not quote data out of context.

Revision and editing principles:
Above all else show the data.
Maximize the data-ink ratio.
Erase non-data-ink.
Erase redundant data-ink.
Revise and edit.

Pugin: “It is all right to decorate construction, but never construct decoration”

Maximize data density and size of data matrix, within reason.

Graphical elegance is often found in simplicity of design and complexity of data.

—

Chartjunk is Tufte’s name for unintentional optical art, such as created by Moire effects. Grids can also be seen as chart junk, because a lot of the time they don’t contribute to understanding, especially when you have to crane your head in all directions to be able to make sense of the written stuff in a graphic.
For grids, he definitely recommends a muted form (grey instead of black, for example), which I’ve used with much success since, although not all figures have made it into the final drafts of papers.

Multifunctioning graphical elements, are, as the name says, elements that convey more than one piece of information. Using these, you can use less elements and still say the same thing. Or you can give redundant information in such a way that you don’t double your elements but try to make everything crystal clear.
To maintain clarity within this chaos, you should try and use graphical methods that can organize and order the flow of graphical information. So no pie charts! They aren’t useful, the information you can put in there is usually easier understood when in a table or in a form where you actually can compare and contrast small-sets of data. Techniques that might prove useful are multiple depths for viewing and multiple angles for viewing (although I’m not sure how to achieve this yet). One thing is for certain, you should succeed in showing the architecture of your data to the viewer.

Another way to see if you are actually showing a lot of information is by looking at data density. Data density can take advantage of the fact that the human eye can detect a large amounts of information in small spaces (although I’m not sure how applicable this is when using a beamer, or a low-resolution screen, considering that paper still has more possible options to put more ink in small spaces than a screen).
A formula to use for data density is to divide # of entries in the data matrix by the area of the data graphic.

A very practical way of using data density is by shrinking your graphics, and portraying them into small multiples. Especially when you want to compare graphics this can be very effective, while using the same template for every graphic is also very efficient. It can communicate narrative content also very well. So far I haven’t really used this optimally, because I was worried for data labels. I’ll putter around with this in the near future and if I come up with something useful show it here. MPhD (co-student) has done some useful stuff with this, to compare results from simulations which are graphs without names or necessary numbers.

However! Sometimes, a super-table can also be useful as a reference, especially when it is easier to read information rather than look at a 100 little bar charts. Elements need for super-tables are organized and sequential detail & a reference-like quality.

Now, on the integration of data and text. Anybody who’s ever written anything in Word knows how difficult this can be. I’ve used Adobe InDesign for lay-out as well, and it’s such a joy when you put your figures in and they don’t move, no matter what you do to the text. In effect, it should be possible to have your figure next to the paragraph where you’re talking about it. The graphic is part of the paragraph!
One thing Tufte cautions about is the difference in use for graphics. You can use graphics to communicate and illustrate a finding, and then you can tell viewers what they are reading. If you’re showing an exploration of your data though, only tell them how to read your design because they will want to make up their own mind. This is very much in keeping with academic tradition, as I see it. Give people the space and time to question and be critical, how else will you learn to appreciate the space of their brain?

Then there’s the proportion and scale issue. How large? Conic? Horizontal? Circle? Tufte proposes to give graphics a horizontal slant, so give your graphics greater length and lesser height (humans are very comfortable with divisions around the Golden Section; so as a rule of thumb you could think about moving towards graphics that are 50% wider than that they are tall). Another point is that cause is often on the x-ass and the effect on the y-ass. Might as well exploit that! Although I have to have a critical look at how I can achieve this every time.