3
$\begingroup$

Are there other ways to represent time series part-to-whole data instead of using an area chart like the one in the example below?

I want to highlight changes in population year to year. And I have some folks that find this a little confusing.

enter image description here

$\endgroup$
2
  • $\begingroup$ Naturally there are other ways to do this. One is show fractions rather than amounts. Another is to show each time series separately rather than cumulatively. Another is to use a transformed scale. These methods can be combined. $\endgroup$
    – Nick Cox
    Dec 12 '13 at 11:46
  • $\begingroup$ You'll get better responses if you can include that or a typical data set in your question (either inline or an external link). $\endgroup$
    – xan
    Dec 12 '13 at 16:29
2
$\begingroup$

I thought* the article by Stephen Few1 interesting to read, so I would like to share one solution he provided there.

He starts his article citing the most three traditional ways to plot time series and part-to-whole data:

  1. sequential pie charts weighted in size by total amount (page 2);
  2. stacked bar chart (page 2);
  3. stacked area chart (page 3), which was the type showed in this question.

The author states the better to visualize total amounts and trends is #3. However, he emphasizes the constraint that the audience needs to be aware about how to read stacked area plot. See the article to visualize disadvantages of #1 and #2.

Then, he provides some alternatives to plot time series and part-to-whole data.

One of them is to present data in two (or more) charts rather than one. The first one will focus on visualize absolute totals and its trend without caring with categories relationships.

I made up some data:

enter image description here

and then, it can be presented the categories chart beneath the total, but with rescaled y axis to better emphasize the relationship among categories.

enter image description here

There is also the option to add a third graph showing categories by percentage of total.

If the category plot is too clutter, the author advises the possibility to stratify the categories according to some sort of criterion (when it is possible). Something like below:

enter image description here

He supports this approach with five arguments cited on the end of page 05.


Reference

1 - FEW, S. Quantitative Displays for Combining Time-Series and Part-to-Whole Relationships. Perceptual Edge. Visual Business Intelligence Newsletter. 2011


*I believe you were also the OP (lopez235) from this thread, right?

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.