1
$\begingroup$

I have two timeseries distributions: you can basically consider them as new tasks and completed tasks. In the standard case, you would try to complete the tasks as and when they come, meaning that the two distributions are near-identical.

I want something that can work well in both these situations. If you graph these as lineplots, you'll basically see just one line in the ideal scenario, which can be confusing - at a glance, you might think that you're only receiving new tasks right now without processing any, or that you're only processing tasks but not receiving any new ones. On the flip side, if the two lines don't line up perfectly, that indicates a problem - and is very easy to spot.

The best idea I have is to make new tasks an area graph, and completed tasks a line graph. However, I feel like spotting the "problematic" parts of the graph is much harder than in the two-line-graphs case, because you'll have to look for a gap between the area and the line. Also, I feel the area graph might just not be too intuitive. Making one a bar graph makes it even harder to spot the issues, stacking the graphs in anyway makes it hard to understand. Graphing only the difference takes away important data, and I'd prefer having fewer graphs.

Does anyone have other suggestions? (Limitations of the graphing software are not a concern, except it has to be a time vs value graph)

$\endgroup$
2
  • $\begingroup$ Please explain further what is the x-axis and y-axis of your intended plot exactly. The wording is still somewhat ambiguous. $\endgroup$ – Nuclear03020704 Jul 19 '20 at 9:28
  • $\begingroup$ x-axis is time, y-axis could be the number of tasks of each type (new/completed), non-cumulative. The y-axis can be changed, though, if it can give a better viz. The only restriction is that the x-axis is time. $\endgroup$ – notablytipsy Jul 19 '20 at 9:30
1
$\begingroup$

This is actually an interesting problem. I can give you some ideas:

  1. Make the line on the back slightly thicker/wider. That is, when they line up you still can see both lines, but when they don't you still can see them not lining up. The drawback is when the difference is too slight, the two lines will seem to still line up as the thin line will not always move to the outside the thick line. Or,

  2. Aside from only two line plots, consider adding an error bar on each points, referencing to the new task. But exaggerate the error bar so even the slightest difference is clearly shown and can be noticed. No error bar means the two line up. This way you visualize difference while not losing count value. Or,

  3. Use your current implementation, but add a label for each point. I assume that new task will always be above completed task, or at least for most of the time. Label new task using a number above the line, and label completed task below the line, therefore you can always see if the two numbers match. Or just style new task boldface font and the completed task italic font.

Hopefully this helps you visualizing your data.

$\endgroup$
3
  • $\begingroup$ Hmm the error bar idea is tempting; I might experiment with that. Thanks! $\endgroup$ – notablytipsy Jul 23 '20 at 21:34
  • $\begingroup$ You can opt to accept this answer if it helps you. $\endgroup$ – Nuclear03020704 Aug 10 '20 at 16:15
  • $\begingroup$ hey, sorry - forgot to get back to you. the error bar idea was a good one, but led to a lot of confusion since it adds an extra line, and people had to be explicitly told that it was scaled up. I'm still looking for something more intuitive. Thank you! $\endgroup$ – notablytipsy Aug 13 '20 at 19:47

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.