I have a set of highly variable datasets, e.g., a line graph for a dataset is attached here. To my eyes, the attached graph is not very readable. I would like to know your opinion on what type of representation might be better for the comparison of highly variable data like what is shown in the below graph.
(1) The use of connecting segments just causes clutter.
(2) A logarithmic scaling for the vertical axis might be an improvement, given that one of the three sets is mostly indistinguishable from zero.
(3) Some indication of what is being measured should be dded, e.g. a label for the vertical axis.
Sometimes it might be possible to re-order x-axis elements (here 'instances') such that the function values of one of the line-plots (methods?) will be non-decreasing/non-increasing. For instance, if you compare the runtime of several methods and you have some baseline, you can re-order the instances such that 'more difficult' instances for the baseline are on the right. Then the graph might become more readable.
Another option would be to aggregate the data somehow - and show just some histograms/bar plots/aggregated characteristics. Maybe, a table of values can be used then instead of a graph.
Also, as noted above, the red line does not give much information, so it might be removed; or the log y-axis might be added.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Graphs