Remarkably easy to learn and use, this statistical software offers many advantages in its price class
click to enlarge
Figure 1: Prism 5 main screen
Prism is statistical software for those analysts needing powerful yet not exhaustive routines. It was originally made for biological studies, but now includes several sophisticated non-linear capabilities as well as scientific graphing that would be useful to other scientists and engineers. The developers have attempted to make this a simple and intuitive exercise and, for the most part, have succeeded. They have attempted to make the learning curve short and gradual and generate results from minimal keystrokes with maximal understanding of the statistical principles being invoked. They have done away with paper manuals and tried to make the on-line help and tutoring as painless as possible. How well did they succeed? Read on…
The software has very modest requirements on the modern computer, needing only 30 MB of hard drive and a minimum of 128 MB of RAM. Version 5 runs on Windows 2000, XP or Vista and on the Macintosh. It will not run on earlier versions of Windows. Screen resolution is 1024 x 768 with at least 16 bit color. Spreadsheet table limits are: unlimited rows by 104 columns. The system downloaded quickly from the internet (“quickly” being dependent upon one’s computer and Internet connect) and installed with no problems on the Vista system.
As there are no manuals, the GraphPad folks have supported the new user with an extensive and easy-to-search Help menu, an on-line FAQ and problem database, a really nice tutorial (either as a video or written text), and an e-mail tech support address. It is unfortunate that there is no phone support, as both novice and experienced user can get maximal mileage from one well-researched phone call. The e-mail turnaround, however, is wonderfully rapid (less than one hour on my first attempt). The introductory tutorial is strongly recommended and will have the first-time user up and going within an hour. The search function within the Help menu was refreshingly sophisticated, and the use of quotes will find only those word strings containing everything between the quotes. Too many otherwise excellent statistical programs lack this function and have very poor indexing. The program is amazingly intuitive and not at all hard to learn on-the-fly.
Although it is relatively easy-to-use a program with cooked data examples, the analysts’ first major problem is data importation and the first headache concerns data manipulation. Prism has several features which greatly simplify the former and perform a few manipulations that will make the latter a bit less painful. Upon first boot-up, Prism 5 displays its main screen (Figure 1). This includes the toolbars, spreadsheet, navigator area and the ‘New Table & Graph’ dialog box.
The Navigator area to the left is usually blank if no previous work had been done, and toolbars are set both above and below the main screen. The top menu bar is rather large and, although suggesting the intuitive ease of performing an analysis, belies the actual breadth of possible analytic routines. Now, as to the data importation and manipulation steps…
Data importation and manipulation
The first big surprise is that, by choosing the proper format for the desired analysis on this dialog box and clicking on the choices for the existing format, Prism will automatically choose the setup
click to enlarge
Figure 2: An annotated graph
correctly for this type of analysis. Choices include XY, Column, Grouped, Contingency and Survival. When this choice is selected, a submenu of tests appears specific to this type of analysis (e.g., regression, correlation, nonlinear regression, curve smoothing and area under the curve for the XY choice). There is a convenient Table Format button available to quickly change the format or number of sub columns. Missing data points are recognized by the software and treated as such. There is an added “sticky note” type of feature that allows the analyst to append notes to each data set detailing its format and analysis. In many programs, this feature is buried under a File/Properties area, and I found this new method to be far more convenient.
Data is easily imported into Prism using either File/Import, or the Import button. While importing data, the user can easily filter, transpose or select incoming rows and columns. Excel is readily imported as either a .csv or .txt file (Prism seems to prefer the .txt) and the Help section recommends a simple cut and paste. This brings up one of the more important shortcomings of this program. Obviously, cut and paste is fairly simple for small data sets, but for large data sets where only a portion of the total is used, can become more cumbersome. Here, the take-home point is that Prism is made for the smaller data sets. Most applications in many disciplines are easily accommodated, but for the newer instances in areas such as molecular biology (specifically genomics or proteomics) the standard sets are far too large for Prism. Now for the really nifty features…
Once data is entered, Prism will automatically create the requested graph. This graph can be quickly modified by an extensive menu of choices that are available from right-clicking on the graphic. This includes the usual symbol
click to enlarge
Figure 3: Nonlinear regression menu
shape/size/color options as well as background colors (very useful in emphasizing the graphic) and the addition of text boxes, Greek letters and formulas. In addition, right clicking on a symbol will allow a change to only that symbol. There is a resize button that will speed the process, but seemingly no click-and-drag feature to immediately resize. However, the software has a MAGIC (Make Graphs Consistent) button that instantly changes a series of graphs to identical formats. As Prism will automatically link the graphic to a data table and any analyses performed, you can make changes in the data that will be quickly updated in the graphs and analyses. To further annotate the graph with pertinent information, portions of the analysis table can be cut and pasted onto the graph (Figure 2).
These graphs may then be exported as wmf, emf, pdf, eps, tif, jpg, png, bmp or pcx formats for publication or exportation to other programs. A single click will send them to a Word or PowerPoint file. Once a graphic contains all of the pertinent information in a favored format, the analyst can clone the graphic, which copies all of the features of the graph to any new data set. Using the Layout feature allows multiple graphs on a single sheet. It is presently not possible to do 3-D or rotatable graphics in this program.
click to enlarge
Figure 4: Classic equations submenu
As is now obvious, Prism is a graph-centric program that likes to visually represent data in quick and easy terms. This, however, shorts the analytic capabilities which, while not extensive (the software does not do prediction/classification problems and multivariate analyses except for 2-way ANOVA), will satisfy many users. Table 1 lists most of the testing capabilities. An added benefit is the priceless (for the non-statistician) Analysis Checklist provided by the Interpret button. This provides an invaluable review of the basis of the tests, conditions that must be met before performing the test in the context of interpreting the results. These explanations are usually very clear and avoid the equations and statistical jargon that quickly bury the novice. Most software nowadays contains helps such as Wizards and mouse-overs to provide context to an analysis, and this is a really refreshing addition to the list.
Also note that, under any given general test scheme, there are many sub-choices, as can be seen for nonlinear regression in Figure 3 and each boxed plus sign lists more sub-choices again (Figure 4). This gives more of an idea of the true menu size.
Prism also keeps track of everything done and remembers all logical links between data tables, sheets, results tables, graphs and layouts. Updating from changes is instant.
From the above, it is readily apparent that Prism was originally designed for biochemical and biological studies, but may be profitably used in a number of areas. The software is remarkably easy to learn and use, and offers many advantages and few drawbacks in its price class. My only regret is that it is not automatically bundled with Harvey Motulsky’s excellent introduction to Bayesian clinical analysis (cleverly disguised as introductory biostatistics), an eminently readable treatise for the beginner .1
A 30-day, full-featured demo is available at the Web site and can be easily put through many elementary exercises within a half-hour.
1. Motulsky, H. Intuitive Biostatistics. Oxford University Press, New York. 1995.
• $595 commercial
• $450 academic
• network and site licensing available
11452 El Camino Real, #215
San Diego, CA 92130
1-858.259.5770, Fax: 1-858.259.8645
John Wass is a statistician based in Chicago, IL. He may be reached at editor@ScientificComputing.com.
Table 1: Prism 5 Analytic Menu
Area under the curve
Fisher’s exact test
Kaplan-Meier survival analysis
Log rank survival analysis
Odds ratios and relative risk
One-way ANOVA (with post tests)
Repeated measures ANOVA
Wilcoxon-Gehan survival analysis