Skip to content
 

“The best data visualizations should stand on their own”? I don’t think so.

Jimmy pointed me to this blog by Drew Conway on word clouds. I don’t have much to say about Conway’s specifics—word clouds aren’t really my thing, but I’m glad that people are thinking about how to do them better—but I did notice one phrase of his that I’ll dispute. Conway writes

The best data visualizations should stand on their own . . .

I disagree. I prefer the saying, “A picture plus 1000 words is better than two pictures or 2000 words.” That is, I see a positive interaction between words and pictures or, to put it another way, diminishing returns for words or pictures on their own. I don’t have any big theory for this, but I think, when expressed as a joint value function, my idea makes sense. Also, I live this suggestion in my own work. I typically accompany my graphs with long captions and I try to accompany my words with pictures (although I’m not doing it here, because with the software I use, it’s much easier to type more words than to find, scale, and insert images).

7 Comments

  1. Andreas says:

    I will say this: Who ever holds the higher theoretical ground, in my world the customer allways wins the argument. It is my experience that different sectors sees and understands things differently. Some people don't read text at all – and others don't "get" vizualisations. This is also a big point in learning theory and cognition styles.: "People are different". So yeah – do both if you want a lot of people to see the genius dripping from the paper.

  2. Phil says:

    When I "read" a paper, often I only read the abstract and conclusions and look at the figures (including captions), so I like it if everything I need to know — except the details — is included there. This makes me a big fan of long captions that don't just say what is being shown — often obvious from the axis labels anyway — but that also point out interesting or important features and do some of the explication that other writers would put in the text.

  3. Wayne says:

    Andrew, I agree. Text can concisely specify things that would clutter a graphic, while a graphic can convey things that would take too many words to describe. Imagine trying to describe a loess fit through some points with text. Imagine trying to specify several data caveats in a graph.

    I think that text can also be further broken down into two types: "prose" and "equations". Imagine describing, in technical detail, an equation without using mathematical symbols. On the other hand, try giving an intuitive explanation or discuss a story using equations.

    So Pictures, Words, and Equations (for those who understand that language).

  4. statc says:

    <a href="http://www.wolframalpha.com/input/?i=plot+x^2%2B%28y-%28x^2%29^%281%2F3%29%29^2+%3D+1" rel="nofollow"&gt <a href="http://;http://www.wolframalpha.com/input/?i=plot+x^2%2B%28y-%28x^2%29^%281%2F3%29%29^2+%3D+1” target=”_blank”>;http://www.wolframalpha.com/input/?i=plot+x^2%2B%28y-%28x^2%29^%281%2F3%29%29^2+%3D+1
    best wishes to all statisticians!

  5. Karl Broman says:

    Also, if a figure must stand entirely on its own, without explanation, then we can't make anything very complicated or new, and sometimes data deserve a complicated and new visualization, for which we must provide some guide.

  6. Howard Wainer says:

    A picture may be worth 1000 words, but it may take 100 or so words to do it. That is why good labels/captions are critical.

    There are two kinds of good graphs:
    1. A strongly good graph is one that tells you all you need to know just by looking at it.
    2. A weakly good graph is one that tells you all you need to know just by looking at it, once you know what to look at.

    Good labels/captions can transform a weakly good graph into a strongly good one.

    I guess this means I am coming down firmly on Andrew's side of this.

  7. Drew Conway says:

    Andy,

    Thanks for your post, and comment on my statement. In reality, we are in agreement, I think the differences are semantic. I consider the caption to be part of the graphic, and as such can "stand alone" in the sense that this supporting information is useful.

    In fact, the visualization I presented in the post you linked actually has quite a lot of text explanation supporting.

    Likewise, however, there are decreasing returns to scale. If a data visualization actually requires 1,000 words to explain what is happening then it seems obvious that there exists a simpler way of representing the analysis.