I just asked the question How to label a constant y-axis in a grid of plots?

Technically, it is about 'visualizing data', and I expect that cv.SE users would have good answers, but I am not sure if it is actually on topic since it deals specifically with labeling a figure - perhaps it would be more on topic at graphicdesign.SE.


There's overlap, of course. I mean to cast no aspersions on our graphic design site--I wasn't even aware of its existence and I'm sure its users are wise and experienced--but in general I would actively avoid advice from graphic designers concerning the design of statistical graphics unless they had statistical credentials as well. And even then I would be cautious.

(My favorite example of how far wrong you can go is Bounford's Digital Diagrams book. Written by a well-established graphic designer and promoting itself as "effective design and presentation of statistical information," this book consistently provides hundreds of examples of everything Cleveland, Tufte, Wainer, et al. have shown to be truly awful approaches: an atlas of chartjunk and other distracting embellishments of all kinds guaranteed to obscure and warp any information that might have been present in the data.)

Thus, when questions come along even about seemingly trivial elements of the design of statistical (or scientific) visualizations, I believe they are very much on topic here.

enter image description here

  • 6
    $\begingroup$ what a great reference! I couldn't resist pasting a shot of the back cover into your answer - I hope you don't mind, but I think that the approach is effective in this particular context (that is, to illustrate your point). $\endgroup$ May 23 '12 at 21:49
  • $\begingroup$ I posted a new question that has brought up the issue again stats.stackexchange.com/q/29994/1381. It is differently focused enough to provide a good test of what is on topic here. Although based on this post it would seem on topic, it could be a better fit for academia.se. $\endgroup$ Jun 7 '12 at 15:30
  • 1
    $\begingroup$ Thanks for the comment. I moved the question to academia.stackexchange.com/q/1933/258 $\endgroup$ Jun 7 '12 at 16:00

I'm really going to have to agree with @whuber here. I think asking graphic designers who lack much statistical knowledge is to court danger. (Although I will acknowledge that I don't know anything about the SE site, and they may have contributors with statistical expertise.)

What I did want to add is that such a question (albeit not yours, I just checked) may do well on Stack Overflow. Here, we occasionally get questions about how to do something with software (typically in R, but possibly using other languages) that have little actual statistical content. If the question is only about how to get software to do something, I think it may be best asked elsewhere. Whereas if the bulk of the question has statistical content, and the software how-to component is the lesser half, I think it primarily belongs here. (Since your question squarely falls in the latter category--it's not about how to get the software to do anything--I think it fits here well.)

  • $\begingroup$ +1 for the clarification in the second paragraph: that's a good statement of our policy. (Concerning the first paragraph, I'm sure it took an effort to agree with me about something: thanks! :-) $\endgroup$
    – whuber Mod
    May 24 '12 at 14:04

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .