Plotting

Plotting your results is a key component of your research. Plotting is writing. Writing is rewriting. Thus, plotting is replotting.

While preparing a paper, expect to revise figures many many times!

When people read a paper, they will tend to read title, abstract, then figures. So your figures should be the “plot” of your paper.

Big picture points

• Most plots will be static representations of data
• The plots in your SI material must be just as high-quality as the main plots.
• Your plots based on data should be reproducible! This means that they should not be touched by Photoshop/Illustrator/Inkscape.

That last point re: Photoshop et al. may be somewhat controversial. A lot of people like to touch up colors, etc., using Illustrator or Inkscape. My opinion is that you are much better off getting the figures right in R or in Python. You should expect to revise figures many times. If your figure work flow involves moving things around in Illustrator, then you’ll be doing that many times, too.

Figure standards

Your figures need to look like figures published in the journals that you read. In your journal clubs, pay special attention to this.

General guidelines:

• No screen shots. Never. Not even for lab meeting. When using output from another application, that application should have a proper way to export results. This goes for Rstudio/Jupyter, too!
• Use proper formatting. If a variable is \(\sigma\), do not write sigma in your plot. Likewise, if you need to write \(\Gamma_{i,j}\), don’t write \(\Gamma ij\) or Gammaij.

Fonts and color deserve special mention.

• Your figure has to be legible when reproduced on a printed page.
• To accomplish this, you need to make the graphic large, with large fonts and high resolution, so that it reduces.
• This takes some time to learn.
• Colors need to be distinguishable more than they need to be appealing.
• Watch out for color schemes that aren’t friendly to common forms of color blindness.
• Any thing like a “heat map” should probably be using the viridis color map. This is the default on Python/matplotlib now, and is an add-on library for R.

Skills to develop

• Saving files in different sizes.
• Saving files in different formats.
• Greek lettering and mathematical formatting.
• Color and alpha transparency.
• Line and point styles/shapes.
• Font sizes: adjusting the axes/legend/title font sizes independently.
• Data-aware labeling. For example, adding an arrow to a plot and printing a number showing the value of the data point. The value should not be hand-written. Rather, it should be a calculation done via code, so that it stays correct if the data change and the plot needs to be regenerated.
• Multi-panel plotting. Ideally, this should be done via code, and not using, say, Illustrator. \(\LaTeX\) users can/should use \subfig environments here, as you can then explicitly \ref each sub-figure.
• Automation. For example, using snakemake to keep plots up to date as the data/scripts change.

Different figures for different contexts

You’ll make figures for exploring data, for lab meeting, for talks, for posters, and for publication. Each of these types will have slightly different needs.

For example, slides for a talk should be much more streamlined than the published version. Many of us fall into a trap of pulling figures directly from papers and putting them into talks. The result is figures with too many extraneous annotations, lines, etc., that distract from the point we are trying to make. (In the paper, the caption would talk about all that stuff. If the caption does not, then you may consider deleting the annotation.)

So, you may consider setting up your plotting work flow along these lines:

• plot_common.R reads in the data, and does whatever manipulation is needed to get the data “plot ready”.
• plot_labmeeting.R, plot_poster.R, etc., may make slightly different versions of the figure.

The contents of the latter may look like:

# do the main processing
source("plot_common.R")

# load up your plotting library
library(ggplot2)

# Do the plot here

Resources

• This is an excellent book on making nice plots based on data. It is written using R, but the concepts apply to all languages.

What else?

What plotting things do you have questions about? I may be able to add them into the week’s exercise.