Minireview: GraphPad Prism for OS X

GraphPad Prism 5
Manufacturer: GraphPad Software
Price: varies depending upon license

Just about anyone who works in science knows that you can have all the data you like, but if you can't show statistically significant differences between groups, you haven't shown anything. And just about anyone who has taken a stats course at university can tell you that calculating statistical significance is both hard and boring. Back before the widespread use of computers in research, it was tough luck—you had better know how to calculate t tests or regressions. But in a perfect example of letting the silicon chip do the heavy lifting, programs were written that would calculate those for you.

Of course, back in those days, everything was harder. There was no PubMed, and definitely no Papers, so searching the literature involved hours spent at a desk in the library poring over Index Medicus. If you wanted to graph something, you needed a ruler, a sharp pencil, and some Letraset for the labels, and that's without touching on having to use log scales and drawing curves by hand.

Luckily for me, I started my career in science in the 1990s, so things like logarithmic graph paper and slide rules were only mentioned by old-timers at the pub. Instead, I was introduced to a software program by the name of GraphPad Prism that not only produced great-looking graphs, but also incorporated a statistics package that was both powerful and, more importantly for me, easy to use. Instead of spending hours calculating a test, you could select the data, choose your parameters, click a button, and voila! GraphPad was founded by Harvey Motulsky, a researcher who was looking to analyze his own data, and I think I speak for most of my peers when I say thank goodness he did!

All those years ago, that first introduction to Prism 3 was on a Windows machine. Shortly afterwards, I made the switch to a Mac, safe in the knowledge that there was a Macintosh version—Prism and I have been good friends ever since. Now the software is at version 5, following a comprehensive rewrite in Cocoa, and is now a universal binary. I'm here to tell you it's been worth the wait.

Each file now contains its own toolbars 

Unlike previous versions, which featured a static toolbar and then document windows, Prism 5 adopts the new format of having each project contain its own toolbar. This might not sound like much, but if you have multiple monitors, it's a massive boon.

The data in each file is separated into five different categories: Data Tables, Info, Results, Graphs, and Layouts. The first of these, Data Tables, is similar in first appearance to a spreadsheet, with rows and columns, and it's into here that your raw numbers go. How the data table is set up depends upon what kind of graph you intend to make with it; is it a simple XY graph, or means and standard deviations for a bar chart, or perhaps a scatter plot? Data can be input by hand, but these days most bits of lab equipment you'll find are nice enough to output their results as CSV files or the like, which you can either import or just copy and paste in.

Like Excel or Numbers, you can manipulate this data by transposing rows and columns, inserting missing values, excluding individual cells and so on. But the real power comes with the Analyze button. You can perform a range of analyses on your data, from simple transformations (X=LogX, X=K*X and so on) on to more complex statistical analyses ranging from column stats through t tests, ANOVAs, and on to linear and nonlinear regressions, correlations and survival curves. The results from these analyses appear in the Results section, and it's here where you find out if that experiment you've been slaving over for the past few weeks actually worked, or if your P value is greater than 0.05 and it's time to go back to the drawing board.

The software also comes with a statistics guide written by Prism's creator, and I've found this guide to be invaluable when it comes to determining exactly which stats test I ought to be using. That sounds like a small thing, but you'd be surprised at the number of researchers I've come across who have no idea when or why to use an ANOVA over a t test, and it's something that often catches people out when it comes to peer review.

The Info section allows you to enter notes on your data: who you are, what the experiments in that file consist of, the date, and so on. It sounds like a minor point, but If you've got 20 or 30 seemingly identical-looking files, a few keywords and Spotlight make it child's play to find the one you want. With this newest version, your graphs will also appear in QuickLook (assuming you're using Leopard), which in practice is a brilliant way to dig through folders of files and pick the ones you want.

Below Results are all your graphs, and you'll quickly recognize the style from most of the papers you read. Simply put, the graphs you can make in Prism are much better looking than those from something like Excel, and you have an awful lot of control over the size and shape and layouts, including things like split axes, custom ticks, angled labels etc. 

Prism offers you a number of default color schemes, but it's very simple to create new ones, and one of the coolest new features in Prism 5 is the ability to make all the graphs in a project consistent. That might not sound like much, but when you've got 15 or more individual graphs in a file and your Principal Investigator tells you he wants them all to look slightly different, being able to make those changes in a minute as opposed to an hour is a life saver.

Graphs can be arranged in layouts of multiple graphs, and both graphs and layouts can be exported to all of the image formats you would want. They can also just be pasted straight into Word, Pages, Keynote or PowerPoint.

Another new feature in Prism 5, one that everyone seems to be doing now, is the inclusion of a new file format that makes use of XML. Unlike Microsoft, GraphPad has had the common sense not to make the new file format, .pzfx, the default, since users with prior versions would not be able to open them. Are you listening, Redmond?

So there you have it. If you're working in biomedical research and you need to analyze your data, you ought to look at Prism (and if you're working in biomedical research and you don't have to analyze your data, something might be wrong). I made a comment in the Office 2008 preview that it matters whether or not software is pleasant and easy to use; life is too short for badly-written apps that you need to fight to get the best out of. Prism is not one of the latter. Rather, it does what it's supposed to, and does it well. It may be somewhat specific to biomedical research, but then those fields do still comprise the bulk of the science done in the US and other developed nations today; compare the NIH's budget to that of the NSF if you're in any doubt. While there might be other stats programs and graphing apps out there, I don't know of any that are as user-friendly nor as powerful. Oh, and there's also a Windows version for those of you who like your mice to come with two buttons.