0

I'm trying to find an alternative way to visualize this data that may be more visually appealing and meaningful, but I'm not sure what other visualization I could use.

Each line represents a test run of different data sets, so I cannot lump them all together. The % Error is the deviation from the mean of multiple monte carlo simulations for each data set, and the % Uncertainty is the 1-sigma standard deviation of the multiple monte carlo simulations for each data set. The variable $x$ is the simulated singnal-to-noise for each monte carlo test.

Is this the best I can do for visualizing this kind of data? My limitations are essentially the size of this plot. I cannot add more plots per journal guidelines.

Edit: this question address a similar problem, however most of these solutions require the use of multiple subplots. While these are fine solutions, I am specifically looking for solutions that don't involve a large number of subplots.

enter image description here

maelstromscientist
  • 101
  • 1
  • 3
    What message are you trying to convey with your visualization? Articulating a clear objective is the first consideration in any graphical design. Characterizing the intended audience is also high up there. Good solutions will vary radically according to these criteria. – whuber Sep 04 '20 at 15:58
  • Does this answer your question? [Visualising many variables in one plot](https://stats.stackexchange.com/questions/190152/visualising-many-variables-in-one-plot) – Nick Cox Sep 04 '20 at 15:58
  • @whuber These are performance measures for an algorithm, so the target audience is anybody who is interested in how well the algorithm performs as a function of x, notably, astronomers. – maelstromscientist Sep 04 '20 at 16:01
  • @NickCox Unfortunately not, as these solutions involve making more plots. I could always use more plots, for each data set, and use something like a box plot, but the problem here is multiple datasets within one (or two in this case) plots – maelstromscientist Sep 04 '20 at 16:04
  • 1
    You can make each panel small and use the same total space. Otherwise: plot 12 series superimposed, plot 12 series juxtaposed, do both; there aren't any other choices that I know of unless you want to reduce the data to a summary or envelope. – Nick Cox Sep 04 '20 at 16:09
  • @NickCox I should explain why I do not specifically like the multiple subplots method. I will also be showing other similar plots as functions of y, and z, with similar numbers of tests. Using multiple subplots for each of these would require a page full of subplots which could look overwhelming for the reader. It doesn't quite provide a more elegant or meaningful solution to the problem, it only splits up the tests into multiple subplots. But thank you for your suggestions. – maelstromscientist Sep 04 '20 at 17:00
  • I think your criteria lead you back to your existing design as the only possible solution for you. – Nick Cox Sep 04 '20 at 17:11

0 Answers0