What does "publication-quality plots" mean?

Offtopic
1

I’ve seen this term many times here and each time got a feeling I live in a different world, where there is nothing special for a plot to be published. So I just took a small random sample of journal articles from my Zotero collection and looked what type of plots are used. The sample happened to include almost all natural sciences (except Astronomy and Geosciences), Mathematics and Engineering. Of 15 papers, only 2 contain plots other then the basic line / scatter / column types.

So what is that different with me? Is it natural vs. unnatural sciences? :wink:

Pub. No. Year Math. Physics Mat. Sci. Chem. Biochem. Biol. Eng. Line 3D-Line Line/ Scatt. Scatt. Column Heatmap Cont. plot Other
[1 ] 2009 + + + 7 1
[2 ] 1993 + + + 1 5
[3 ] 2015 + + + + 1 6
[4 ] 2006 + + 4 5 1
[5 ] 1991 + 4
[6 ] 2002 + + 1 10
[7 ] 2004 + 3 9 1
[8 ] 2005 + + 4 2 2
[9 ] 1987 + + 1 1
[10 ] 2018 + + 5 1 1
[11 ] 2000 + + 1
[12 ] 2009 + + 1 1
[13 ] 2006 + + 1 1 2
[14 ] 2008 + + 3
[15 ] 2001 + + +
1 Like
2

In my view, a “publication-quality plot” has absolutely nothing to do with the inside contents of the plot. It’s all about workflows. In short, I’d say a “publication-quality plot” is simply one that does not need post-processing in a graphics editor to make you and your reviewers happy.

It’s about getting the font-sizes right and the labels readable. It’s about fine-tuning the positioning. It’s about axis styles and colors and preferences. It’s about doing this programmatically so I don’t have to dive into Illustrator/Inkscape/whatever every time I re-run my analysis.

11 Likes
3

For me it means:

  • vector graphics
  • LaTeX can be used for any text
  • free choice of the fonts to be used (often publishers
    require that the same font is used for the main text
    as in the diagrams)
  • often black and white is the better choice, so you
    need a large number of line styles
  • it must be possible to scale the graphics and the text
    separately
  • logarithmic scaling in good quality

Simple example (this is not vector graphic because Discourse does not allow it,
normally it would be .pdf)
Noise

9 Likes
4

I’ll agree that “publication quality plots” is a bit vague - for too many authors, this can mean “screenshot from Excel.”

Perhaps it means: “not hand-drawn” (unless you are a good artist)

5

I totally second that. It is really about readability and scalability for me, so definitely vector graphics if possible. When exploiting experimental data, it might be difficult to use pure vector graphics but, I spent a bit of time combining PGFPlots axes and images and the result is surely better than having rasterized labels.