Graphs

Joseph Green, DMSc

Two rules of good graphs are presented and explicated. (CHEST 2006; 130:620-621)

Key words: graphs; medical literature; tables; writing

I am at a loss for words. I did not expect that looking through recent issues of CHEST would cause both relief and disappointment. Relief, because the topic this month is still relevant. The need for advice about graphs is as strong now as at any time in at least 2 decades. Disappointment, for the same reason: newly published medical literature still has far too many graphs that are unnecessary, distracting, difficult to decipher, and sometimes even deceptive.

At the risk of oversimplification, I suggest that there are two rules of good graphs. First, a good graph shows data that deserve to be graphed. It displays or reveals relations or patterns that would be hard to grasp if the data were shown in running text or in a table. Second, a good graph succinctly emphasizes the data, not the graph itself. Just as good writing has no needless words,1 good graphs have no distracting lines or other needless marks of any kind.

To see how these two rules work, let us inspect a graph published in a medical journal that shall remain nameless, although I will say that its impact factor is 5 to 10 times higher than that of CHEST. That fact is not trivial. Some of the worst graphs can be found in the most prestigious journals. I redrew the graph as faithfully as I could, although I changed the labels to avoid violating copyright. (Why a publisher might want to protect its copyright to junk is a question I leave to others.) The offending graph is shown in Figure 1. It reminds me of a bit of wisdom I learned from my mother: Es genügt nicht, keine Gedanken zu haben. Man muβ auch unfähig sein, sie auszudrücken. (It is not sufficient to have no ideas. You must also be incapable of expressing them.) Not only do the data shown in Figure 1 have no right to be shown in a graph, they are shown in a graph that is badly drawn.

What is the first thing you notice about Figure 1? Perhaps it is the fact that this graph shows only four values (four data). Good graphs showing only four values are not unheard of, but they are uncommon, and this one is not among them. What relationship or pattern is revealed in Figure 1 that could not be described at least as well in words or shown at least as clearly in a two-by-two contingency table? None. The first rule of good graphs has been violated. These data should not have appeared in a graph at all.

Figure 1. A really terrible graph of four values
that should not be shown in a graph at all.

But wait: there is more. Not only does showing those data as a graph add nothing to the reader’s understanding, it also gives the readers who want to comprehend the author’s point a task that is unnecessarily difficult. It does so by distracting us and thus violating the second rule of good graphs. The heights of the columns tell us something about the values they are intended to represent, so we need not ask why the columns are as tall as they are. So far, so good. Now, what purpose is served by making the columns so wide and so deep? None. None, that is, unless the author intends to distract, or to appear important, or to waste paper. But would making the columns narrower and shallower turn this sow’s ear into a silk purse? No. We (at least some of us) would still be distracted by the phony third dimension. It is completely unnecessary. Even the use of two dimensions to show one value can be reasonably questioned; the phony third dimension is overkill.

I wish that you, as a contributor to CHEST, could sit back and say in all sincerity “I don’t make graphs like that!” But many of you do. Graphs of data that could be shown just as well or better in running text or in tables, and graphs that distract the reader’s attention rather than directing it appropriately are easy to find, even in CHEST. Fortunately, they are also easy not to make. As someone who lives in a glass house but still throws stones, I recommend the book by Mary Briscoe2 (still in print after 10 years), and the section on graphs in the second edition of How to Report Statistics in Medicine by Thomas Lang and Michelle Secic (which should be available by October 2006).3 If you want to learn more about graphs, I recommend the works of Edward Tufte,4-6 William Cleveland,7,8 and Howard Wainer.9

Even without a second opinion, you can make progress on your own. Be honest about why you think your data should be shown in a graph and, if you still choose to make one, ask yourself whether your graph draws attention merely to itself, or to important relations and patterns in the data.

References

1 Strunk W, White EB. The elements of style, 3rd ed. Boston, MA: Allyn and Bacon, 1979

2 Briscoe MH. Preparing scientific illustrations: a guide to better posters, presentations, and publications. 2nd ed. New York, NY: Springer-Verlag, 1996

3 Lang TA, Secic M. How to report statistics in medicine: annotated guidelines for authors, editors, and reviewers. 2nd ed. Philadelphia, PA: American College of Physicians, 2006

4 Tufte ER. The visual display of quantitative information. Cheshire, CT: Graphics Press, 1983

5 Tufte ER. Visual and statistical thinking: displays of evidence for making decisions. Cheshire, CT: Graphics Press, 1997

6 Tufte ER. Visual explanations: images and quantities, evidence and narrative. Cheshire, CT: Graphics Press, 1997

7 Cleveland WS. Visualizing data. Summit, NJ: Hobart Press, 1993

8 Cleveland WS. The elements of graphing data. Summit, NJ: Hobart Press, 1994

9 Wainer H. Graphic discovery: a trout in the milk and other visual adventures. Princeton, NJ: Princeton University Press, 2005

*From the Graduate School of Medicine, University of Tokyo, Tokyo, Japan.
Correspondence to: Joseph Green, DMSc, University of Tokyo,
7-3-1 Hongo, Bunkyo-Ku, Tokyo 113, Japan; e-mail: jgreen@m.u-tokyo.ac.jp
DOI: 10.1378/chest.130.2.620

グラフ

Joseph Green, DMSc

キーワード： graphs; medical literature; tables; writing

そんなに単純ではないといわれるのを覚悟で提言するが、よいグラフには2つのルールがある。第1に、よいグラフはグラフ化するに相応しいデータを表している。つまり、文章や表にすると把握しづらくなる関係やパターンを示しているのである。第2に、よいグラフはグラフそのものでなく、データを簡潔に表している。よい文章に不必要な言葉がない1のと同じように、よいグラフには注意をそらすような線や不必要な印がない。

この2つのルールが実際にどう適用されているかを理解するために、ある医学学術誌に掲載されたあるグラフをみていこう。この学術誌の名前は伏せておくが、そのインパクト・ファクターはCHESTの5倍から10倍高い。この事実は大きな問題だ。最悪のグラフが一流の学術誌に掲載されていることがあるのだ。グラフはできるだけ忠実に再現したが、著作権を侵害しないようにラベルを変更した。（なぜ出版者はくだらないものに著作権を保護したがるのか？という疑問は、どなたかにお任せする。）問題のグラフは図1に示されている。これを見ると母が教えてくれたこんな言葉を思い出す。Es gen?gt nicht, keine Gedanken zu haben. Man muβ auch unf?hig sein, sie auszedr?cken (アイディアが何もない、というだけでは不十分。何もないアイディアを表現する能力も備わっていてはいけない)。図1に表されているデータはグラフにするのに適切でないだけでなく、グラフそのものもよくない。

4つの値を表した非常に劣悪なグラフ

しかし、ちょっと待った。これでおしまいではない。このデータをグラフで表すことは、読者が理解する何の助けにもなっていないだけではなく、著者が伝えたい要点を把握することを必要以上に難しくしている。このグラフは読者の注意をそらしている。従ってよいグラフの第2ルールが破られている。棒の高さには、何らかの理由がある。だから、なぜ高いのかなどということを聞く必要はない。ここまではよいとして、では、棒をこんなにも幅広く、奥行き深くする目的は何なのか？答えは「何もない」だ。でなければ、著者はわざと「読者の注意をそらそうとしている」か「自分を重要に見せようとしている」か、あるいは「紙を無駄にしようとしている」。しかし棒を細く、奥行きを浅くすれば瓜の蔓に茄子がなるのか？「ノー」だ。意味のない三次元は、我々（少なくとも何人か）にとっては、やはり注意をそらされる。これはまったく不必要である。1つの値を示すのに二次元を使用することさえかなりの疑問であるのに、意味のない三次元はやり過ぎだ。

セカンド・オピニオンを求めるまでもなく、読者は自分で改善することができる。なぜデータをグラフにする必要があるのかを素直に考えてほしい。それでもグラフを作ると決めたら、注目されるのはグラフそのものだけなのか、それともデータが表す重要な関係やパターンなのかを自問することだ。

1. Strunk W, White EB. The elements of style, 3rd ed. Boston, MA: Allyn and Bacon, 1979

2. Briscoe MH. Preparing scientific illustrations: a guide to better posters, presentations, and publications. 2nd ed. New York, NY: Springer-Verlag, 1996

3. Lang TA, Secic M. How to report statistics in medicine: annotated guidelines for authors, editors, and reviewers. 2nd ed. Philadelphia, PA: American College of Physicians, 2006

4. Tufte ER. The visual display of quantitative information. Cheshire, CT: Graphics Press, 1983

5. Tufte ER. Visual and statistical thinking: displays of evidence for making decisions. Cheshire, CT: Graphics Press, 1997

6. Tufte ER. Visual explanations: images and quantities, evidence and narrative. Cheshire, CT: Graphics Press, 1997

7. Cleveland WS. Visualizing data. Summit, NJ: Hobart Press, 1993

8. Cleveland WS. The elements of graphing data. Summit, NJ: Hobart Press, 1994

9. Wainer H. Graphic discovery: a trout in the milk and other visual adventures. Princeton, NJ: Princeton University Press, 2005

（訳）　豊田麻友美 　J. パトリック・バロン