0
$\begingroup$

I want to improve the graphic

enter image description here

that I created with R's ggplot2. The graphic is for a scientific publication. These are the runtime of 12 algorithms. There are four at the top, four in the middle and four at the bottom. They are not necessarily related to each other. It is very hard to discern the different colours from each other.

The main point that I want to drive home is that there are four that are slow, four that are middle fast, and four that are fast. Indeed, the groups should be more or less equally fast, and the fact that they differ is mostly due to the fact, that we talk about runtime (which always differs).

How can I improve the plot such that I drive home the main point without being distracted by the many colour, but possibly still have differentiation within the three groups?

Maybe it is not smart to keep the differentiation in the first place?

I would like to keep the line plot, but I am open to alternative suggestions if there are really good ideas out there.

$\endgroup$

3 Answers 3

1
$\begingroup$

You already made it log scale which would have been my first idea. How about use lattice to make it three plots with different Y start and end? Each plot should have a center point that covers one of the three groupings you have and you can stack them 1x3 (vertically).

$\endgroup$
1
  • $\begingroup$ That weakens the statement about comparing the different groups. $\endgroup$
    – Make42
    Sep 21, 2017 at 12:28
0
$\begingroup$

Three things you can try

  • You can add a legend to the plot. Even clearer would be text just above the groups of lines that say "fast", "medium", "slow"

  • Make sure you mention to the audience that it is your subjective grouping, or if you use a metric, to include the definition

  • The loglog plot makes the dramatic runtime differences less pronounced. You can try seeing what this plot looks like with linear axes

$\endgroup$
0
$\begingroup$

If your main objective is to highlight/compare the speed groupings, I think a bar chart might accomplish that more clearly. You could plot 12 distinct bars, showing the total runtime@50% (or whichever % mark you like). The speed groupings would still be clearly shown, and you could label each individual algorithm in the display.

$\endgroup$
1
  • $\begingroup$ But then I would not be able to show the results for all percentages, right? (However, I want to.) $\endgroup$
    – Make42
    Sep 22, 2017 at 13:29

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

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