The zigzag
Someone I work with wanted to know about broken axis graphs (like the one pictured) -- where you put what I call "the zigzag" on the axis. Is there a name for that zigzag? Are there special cases where you use it? How in the name of GTFW do you create one of these things?You've got data that doesn't fill the entire data range. If you show the entire data range, there is a chance that you won't be able to see interesting features the data contains. So I posed the question to the Rhetoric for Engineers mailing list and got the following responses:
Lisa Henn broke out her old drafting book (Giesecke is the author) and found that
[C]ertain circumstances require special consideration to avoid wasted space. For example, if the values to be plotted along one of the axes do not range near zero, a 'break' in the grid may be shown... [W]hen relative amount of change is required... the axes or grid should not be broken, and the zero line should not be omitted. If the absolute amount is the important consideration, the zero line may be omitted.And Lisa herself adds,
On graphs, you'd want to put breaks such that there is still a gap shown in the data. In other words, you wouldn't want the data points crowding the break too much.She finally points out that Edward Tufte probably addresses the issue.
Glen Hadley argues that it's not just that we want to ignore empty parts of the data range, but that we may want the data range to appear fully populated as well. This is an appearance issue, meaning that we might not be interested in an axis break for analysis. It's for presentations. So there are several sites where educators and consultants have found ways to fool Excel into creating axis breaks (why wouldn't Microsoft make that a standard option?) -- this is a long and tedious process, and you can be sure most engineers will never bother with it. I did it myself with the graph above, to make sure I could follow directions, and it took me 20 minutes. I also added a trendline to the original data, something the people writing on this subject didn't add to theirs. Here are the sites where they work this issue:Everett Greene throws in a complaint:
It's too bad the financial world doesn't read this. The economic, stock, and such graphs (almost) always show the top 0.1% of a curve which magnifies the miniscule noise into major ups and downs making it appear that the noise is significant when it isn't.Preach it, brother!
Labels: education, engineering, internet, rhetoric
posted by Ron @ 11:27 AM
Site Feed
0 Comments:
Post a Comment
<< Home