3

What are the recommended guidelines for labeling the axes of an inset displaying a magnified or reduced view of the primary plot? Should an inset include labels if the main plot is in linear scale while the inset is presented in log scale?

Based on my academic experience, individuals frequently add axis labels to insets. Is there a specific rationale behind this practice? I have reviewed several author guidelines but haven't come across any that specifically address this matter.

For an instance of an inset plot (axes are not labeled), please refer to the following example: Inset plots in Matplotlib.

df7
  • 33
  • 4
  • Relevant question: https://academia.stackexchange.com/questions/18357/are-there-any-guidelines-for-labeling-axes-in-plots-graphs – df7 Oct 19 '23 at 04:52
  • 2
    I'm not sure there are any universally accepted guidelines. As long as the graph is clear and easy(ish) for someone in your field to interpret, you're good. – sErISaNo Oct 19 '23 at 05:01
  • 5
    Indeed, take inspiration from figures in journal articles in your field. To me, not labeling axes is a bad practice. – Jon Custer Oct 19 '23 at 14:05
  • 2
    In the liinked plot the inset axis labelling absolutely makes sense. People need to understand precisely what they see, and this helps. – Christian Hennig Oct 19 '23 at 16:59
  • 2
    If the inset and main plots have different scales (log and linear), then the inset plot is not just a magnified or reduced view. And in the linked example, the inset plot is not just a magnified view as it includes a red line that is not in the main plot. – toby544 Oct 20 '23 at 11:40

2 Answers2

3

If you mean the axis labels like T/K in the linked example, then actually I don't think they are necessary in the inset plot. It is obvious what variables are shown on the two axes.

If you mean the numbers next to the tick-marks, then yes those are usually necessary and valuable.

If the inset graph uses a different type of scale for one axis, e.g. log, then it would be better to show it completely separately. Showing it inside the main plot suggests that it at least has the same type of scale, linear or log.

The linked example is not very good in my opinion. It could be improved by making the lines that show the zooming (the little box in the bottom left, the straight lines showing the zooming, and the box around the inset plot) all be bold and in a different colour. And it is not clear why the plot doesn't show the Debye model outside that small range.

toby544
  • 2,561
  • 10
  • 10
2

Every time I peer review a manuscript with an inset plot, the inset is illegible. I have never seen a well designed inset plot.

Change your inset plot to a separate figure. It will be easier to design, easier to read, and the answer to your question will be obvious: Yes, you must always label axes.

Anonymous Physicist
  • 98,828
  • 24
  • 203
  • 351
  • 1
    Especially if you're moving between linear and log scales... – Bryan Krause Oct 20 '23 at 13:54
  • 1
    I forgot to mention I came across an ACS Publications Author Guidelines, which discourages the use of insets. This is mentioned in the "Manuscript Submission Requirements Checklist" under "Artwork Tables/Schemes/ Graphics." – df7 Oct 22 '23 at 07:35
  • Could you please explain what the illegible point would be for Figure 6b in the following paper? https://doi.org/10.1103/PhysRevB.85.035135 It seems to me that the log-log plot in the inset is used to clearly show the appearance of minima. – df7 Nov 09 '23 at 12:17
  • 1
    @df7 That one is definitely better than most, but: 1. The small tics are marginally legible. 2. y axis label kerning is wrong. 3. You can already see the appearance of minima in the big plot. 4. You could see the appearance of minima better if the big plot were not squished to make room for the inset. 5. You gave a reason why the inset was clear; but you didn't have a reason for the big plot. I would probably replace the big plot with the inset. 6. If I read the rest of the paper, will it explain why there are no error bars (include systematic error)? – Anonymous Physicist Nov 09 '23 at 12:55
  • @AnonymousPhysicist I appreciate your thorough clarification of these points, especially the first one. Regarding point#4, it appears to me that the author has maintained consistent scaling for both Figures 6a and 6b. The purpose is most likely a straightforward comparison between two distinct scenarios (H∥b and H⟂b). It's worth noting that the y-axis scaling differs in the case of Figures 6c and 6d. Concerning point#6, the author has not presented any explicit reasoning. – df7 Nov 10 '23 at 00:56