Modern spreadsheet programs offer all kinds of chart options. Most suck.
Dr. Drang recently called out the stacked area chart. I completely agree with the good doctor that stacked area charts do a poor job of exposing trends in data.
Whenever I attempt to present data that vary with time, I always start simple—with a line chart. Yes, line charts are boring. They’ll win you few to nil artistic points.
Consider Dr. Drang’s market share data in the form of a line chart:
No, it’s not obvious that the y-coordinates sum to 100 percent at each time step, but I think that’s secondary to making the trend in the data crystal clear.
I’m no cognitive psychologist, but I think the human mind naturally looks for lines in charts. Stacked area charts have a way of implying lines where they don’t really exist.
Line charts are less apt to create optical illusions because what you see is what you get. The reference point for each line is the same: the horizontal axis. There’s no question that Orange is falling, Blue is rising, and that Green is increasingly linearly.
Because trends are so immediately clear in a line chart, viewers will spend time contemplating the meaning of the trends rather than wasting mental calories decrypting the graph’s basic message. For example, questions like “why are Blue and Orange negatively correlated?” or “why does Green seem independent from the others?” become possible. Busier charts might not elicit those questions.
It’s likely that there is a more creative way to convey the same information, but I think any departure from a simple line chart should be judged against what Edward Tufte would call the line chart’s “data-ink ratio.”
In other words, as you increase your artistic license, ask: Am I adding insight or distraction?
{ 6 comments… read them below or add one }
This articulates exactly why line charts are my choice 99% of the time. Stacked area charts do *look* neat, but Dr D showed a great example of how misleading they can be.
Line charts are boring. But they convey the information needed quickly and clearly. That’s the entire of point of a chart, isn’t it?
Do you know the book “How to lie with statistics” (see http://en.wikipedia.org/wiki/How_to_Lie_with_Statistics)?
It covers the problems you have by analyzing statistic displays and attributing certain diagrams and data a “lie” factor (it often relates to the fact that humans perceive area and “value” differently, e.g. you can display a value using the radius or the area of a circle – no matter what you do, you can use it to your advantage ;)). I haven’t read it but heard of it and saw several interesting parts of it. It might be of interest to you.
As for diagrams, it depends on your data – there’s data that makes sense to look at with a stacked area chart. Picking the right diagram for data is difficult at times and it can lead to many false assumptions as well to much insight in the contrary. Another problem is, that it also depends on the knowledge of the viewer – which may vary a lot ;).
I agree however that most available charts suck in certain ways as they often don’t concentrate on representing data but just looking “good”.
The “data-ink ratio” points to a conspiracy: ink suppliers want to sell more ink.
This article sounds like it was inspired by Tufte.
http://en.wikipedia.org/wiki/Edward_Tufte
His books talk about this kind of stuff at length.
“No, it’s not obvious that the y-coordinates sum to 100 percent at each time step”
Suggestion: help the client by adding a horizontal grid line labeled 100% at the top of the line chart :-)
Putting a 100% at the top does not make it clear that the three values at each point add up to 100, all that would do is show that none of the lines goes to 100. A title like “percentage shares of whole market” is about the only thing you can do.
I’m against the idea of fancy charts in principle. You shouldn’t be looking at a chart for how pretty it is. You should be thinking “why do I need artistic licence for this chart?” before you even think how you should apply it.