15

I've conducted a survey for my masters thesis and would like to show the results as a box plot:

Box plot, as it will appear in the thesis

(The x-axis lists different skill / familarity levels, ranging from "none" to "expert")

Should I mention that this graph is a box plot (in the caption or the surrounding text) so that people unfamiliar with it can look up how to interpret this type of diagram, or is it unnecessary information and I can assume everybody will understand the type of graph just by looking at it?

Context is a computer science masters thesis in German at a German university.

  • 2
    It has a very easy fix: make the X-axis numeric from 1-5, and explain what those numbers mean in the caption, or just simply add "1 - Keine, 2 - ..." as captions. The only reason this plot is confusing is because boxplot are numerical plots and there are no numbers in your plot. – Ander Biguri Jun 4 '18 at 9:14
  • 58
    Since nobody has clearly said it: This is the wrong type of plot for your data. Don't use a Box-plot (nor a violin plot) for discrete data, only for continuous data. – Roland Jun 4 '18 at 10:17
  • 2
    The only fully reliable answer has to come from your advisor. – henning -- reinstate Monica Jun 4 '18 at 11:25
  • 1
    @Roland yeah. Of course one might debate whether the discrete categories make much sense here in the first place – everybody will have a different notion of what “intermediate skills” should mean. – leftaroundabout Jun 4 '18 at 14:14
  • 2
    @Roland: The OP's plot is fine and much clearer than those alternatives. (You're quibbling that the levels in an ordered categorical, and distances between them, don't have much or any numerical meaning, which is a never-ending debate; it depends on how they were measured.) Please don't post absolute claims like "This is the wrong type of plot for your data" if they're not generally agreed. – smci Jun 5 '18 at 9:12
13

In general, yes.

Most potential readers will understand simple graphs (such as box plots).

However, many visualization techniques have variants that produce visually similar graphs but may have slightly different interpretation.

Therefore, it is a better practice of research presentation to clarify what exactly the figure shows and, if needed, how it visualize the study methods or findings.

In your case, for example, in the caption I recommend to mention at least it is a box plot. It is also better to clarify whether the box width is proportional to the group size, in which case the three groups appear to have (almost) the same sizes.

  • 1
    I agree. It is especially important to mention any deviation from norms, i.e. the Whiskers are usually drawn at 1.5 times the interquartile range and outliers beyond this explicitly marked. If the range for the whiskers is different or outliers are supressed, this should definitely be mentioned, else the reader assumes there are no outliers. – Neuneck Jun 5 '18 at 8:36
20

You should use your caption to explain what the graph is supposed to be indicating. i.e. "Boxplot of survey data showing generally higher level of experience with SVN than competing technologies". This achieves both, and further supports the text of your thesis, rather than explaining to the reader something that should be completely obvious (e.g. "This is a boxplot of the survey data"). If you include the type of chart in the caption then it will help if someone is unfamiliar with that type of chart and won't take up much more room than typing "Chart of survey data showing..."

That said (and slightly off-topic), a boxplot doesn't seem to be the correct graph to use to convey this information. A boxplot should be used for continuous data. Here you have discrete (categorical/ordinal) data. Perhaps a barchart (with the different categories shown in different colours might work better)? I'm struggling to understand what the "lower quartile" between "Keine" and "Grundkenntnisse" means for google docs. Is that 25% of respondents had Keine experience of google docs? Very confusing! It seems like you've decided you want to use a boxplot and are trying to find a way to use one.

Let's say I want to know how many people have expert understanding of SVN, how do I determine that from your graph?

However with a bar chart... it's clear that (for example) 5 people who responded were expertsbar chart showing experience of people in different technologies

  • 5
    Answer is off-topic. The question does not mention if the data is continuous or discrete. No need to jump to conclusions. This response is also not useful to people other than the asker. – Anonymous Physicist Jun 4 '18 at 22:20
  • 8
    @Anonymous Physicist. The data seems to be ordinal. Keine = no knowledge, Grundkenntnisse = little, Mittlere = intermediate, Fortgeschritten = advanced, Experte = expert. – henning -- reinstate Monica Jun 5 '18 at 8:06
  • 1
    @henning it will depend on how the question was asked. If the question asked was "On a scale of 1-100 how would you point yourself on this technology" then the plot might be more relevant than you think it is. – John Hamilton Jun 5 '18 at 12:25
10

You probably don't need to mention the word 'Boxplot' in the image caption itself (because it is obvious to the reader; I would assume that most readers will recognize this even if they are not directly familiar with the interpretation of a boxplot).

However if you intend to give a completely unaware reader some hint that what they see is a boxplot, you could also mention the word just once in the paragraph where you refer to the figure (e.g. "compare the boxplot in Fig. 3.2"). I imagine that I would do it this way, but at the same time I don't think that your decision whether or not to mention it would influence the outcome (e.g. grading) in any way.

  • 2
    "hint that what they see is a boxplot" that's it! Just mentioning the name to facilitate the googling of unfamiliar readers. – YYY Jun 4 '18 at 21:32

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.