Early in my career, I wish someone had taken me to one side and explained just how important figures are. Too often I see students fretting over the text, reading endless reams of publications out of concern that they haven’t cited enough, or cited the right things. Or fine-tuning their statistical analyses far beyond the point at which it makes any meaningful difference. And yet when it comes to the figures, they slap something together using default formatting, almost as an afterthought.
Having recently written a textbook (shameless plug), it has only brought home to me how crucial figures are to whether your work will get used and cited*. The entry criterion for a study being used in a book isn’t necessarily the quality of science, volume of data or clarity of expression, though I would argue that all of these are high in the best papers. What really sets a paper apart is its figures. Most of us, when we read papers, look at the pictures, and often make a snap judgement based on those. If the figures are no good then the chances of anyone wading through your prose to pick out the gems of insight will be substantially reduced.
Here then is a useful rule of thumb: you should spend at least one working day preparing each figure in a manuscript. That’s after collecting and analysing the data, and after doing a first-pass inspection of the output. A whole day just fine-tuning and making sure that each final figure is carefully and concisely constructed. You might not do it all in one sitting; you may spend 75% of the time trying out multiple formats before settling on the best one. All this is time well spent. And if you’re going to put the time into preparing them then you should look into bespoke software that will improve the eventual output.
Presenting statistical outputs
If you’ve been following this series of posts then it will come as no shock that I don’t recommend any of Microsoft’s products for scientific data presentation. The default options for figures in Excel are designed for business users and are unsuitable for academic publication. Trying to reformat an Excel figure so that it is of the required quality is a long task, and one that has to be repeated from scratch every time**. Then saving it in the right format for most journals (a .tiff or .eps file) is even less straightforward. As an intermediate option, and for those who wish to remain in Excel, Daniel’s XL plugin is a set of tools for analysis and presentation that improve its functionality for scientists.
Needless to say, this is all easier in R with a few commands and, once you’ve figured it out, you can tweak and repeat with minimal effort (the ggplot2 package is especially good). The additional investment in learning R will be rewarded. In fact, I’d go so far as to say that R is worth the effort for preparing figures alone. No commercial product will offer the same versatility and quality.
One of the reasons I recommend ggplot2 is that it is designed to follow the principles of data presentation outlined in Edward Tufte’s seminal book The Visual Display of Quantitative Information. It’s one of those books that people get evangelical about. It will change the way you think about presenting data, and forms the basis for the better scientific graphing tools.
If you’re not an R user then a good alternative is the trusty gnuplot. Older readers can be forgiven for shedding a nostalgic tear, as this is one of the ancient software tools from the pre-internet age, having been around for about 30 years. It lives on, and has been continually maintained and developed, making it just as useful today as it was then.
Finally, if your main aim is to plot equations, then Octave is a free alternative to the commercial standard MATLAB. Only the most mathematical of biologists will want to use this though.
Some people try to produce diagrams using PowerPoint. No. Don’t do it. They will invariably look rubbish and unprofessional.
For drawing scientific diagrams, the class-leader is the fearsomely expensive Adobe Illustrator. Don’t even consider paying for your own license though because the free Inkscape will do almost everything you’ll ever need, unless you’re a professional graphic designer, in which case someone else is paying. Another free option is sK1 which has even more technical features should you need them. Xara Xtreme may have an awful name but it’s in active development and looks very promising. It’s also worth mentioning LibreOffice Draw, which comes as part of the standard LibreOffice installation.
One interesting tool I’m itching to try is Fiziko, which is a MetaPost script for preparing black-and-white illustrations for textbooks which mimic the appearance of blocky woodcuts or ink drawings. It looks like some effort and experience is required to use it though.
The expensive commercial option is Photoshop, which is so ubiquitous that it has even become its own verb. For most users the free GIMP program will do everything they desire. I also sometimes use ImageMagick for image transformation, but mostly the command-line tool sam2p. Metadata attached to image files can be read and edited with ExifTool.
A common task in manuscripts is to create a simplified vector image, perhaps using a photo as a template. You might need to draw a map, show the structure of an organ or demonstrate an animal’s behaviour. For this there are specialist tools like Blender, Cheetah3D for Mac users or Google’s SketchUp, though the latter only offers a limited version for free download. Incidentally, never use a pixel art program (like Photoshop) to trace an image. All you end up with is a simplified pixel image of the original, which looks terrible. Plus you’ve paid for Photoshop.
For the rather specialised task of cropping and assembling documents from pdf files, briss might be an ancient piece of software but it’s still the go-to application.
Preparing outline maps (e.g. of study sites) is a common task and an expensive platform like ArcGIS is unnecessary. Luckily the free qGIS is almost as good and improving rapidly. There’s a guide to preparing maps here.
Flowcharts, organisational diagrams and other images with connected elements can be created in LibreOffice Draw. I’ve not used it for this though, and therefore can’t compare it effectively to commercial options like OmniGraffle, which is good but expensive for something you might not be doing regularly. A LaTeX-based option such as TikZ is my usual choice, and infinitely better than spending ages trying to get boxes to snap to a grid in Powerpoint. If you’re not planning to put the time into learning LaTeX then this is no help, but add it to the reasons why you might. If anyone knows of a particularly good FOSS solution to this issue then please add in the comments and I will update.
This is one task that R makes very easy. Take the output of a script that creates multiple PNG files from a loop and bundle them into an animation using QuickTime or the very straightforward FFmpeg. For something that looks so impressive, especially in a presentation, it’s surprisingly easy to do.
To collect data from images ImageJ is by far the best program, largely due to the immense number of specialist plug-ins. Some of these have been collected into a spin-off called Fiji, which provides a great set of tools for biologists. Whatever you need to do, someone has almost certainly written a plug-in for it. Note that R can also collect data from images and even interfaces with ImageMagick via the EBimage package. Load JPEGs with the ReadImage package and TIFF files with rtiff.
A common task if you’re redrawing figures, or preparing a meta-analysis, is to extract data from figures. This is especially common when trying to obtain data from papers published before the digital age, or when the authors haven’t put their original data online. For this, Engauge will serve your needs.
Next time: how to prepare presentations!
* At some point in the pre-digital age, maybe in the 90s, I recall an opinion piece by one textbook author making exactly this point. Was it Lawton, Krebs, Southwood… I really can’t remember. If anyone can point me in the right direction then I’d be grateful because I can’t track it down.
** I did overhear one very prominent ecologist once declare only half-jokingly that they stopped listening to talks if they saw someone present an Excel figure because it indicated that the speaker didn’t know what they were doing. Obviously I wouldn’t advocate such an extreme position, but using Excel does send a signal, and it’s not a good one.