Some shocking changes accompany greatly extended analytic capabilities
As the saying goes, ‘Boy will you be surprised!’ From the entirely revamped main screen to the addition of many sophisticated simulations, graphics, an R interface, and the new Degradation platform, this newest version is a major advance. Let’s start with the new main screen which may be a douse of cold water in the face to many experienced users. Rather than the old, familiar spreadsheet and toolbars, we now have a more utilitarian screen (in at least the opinion of the developers) that includes the familiar toolbars, but is dominated by ‘Recent Files’ and ‘Windows List’ areas which occupy most of the screen (Figure 1).
![]() Figure 1: JMP 9 main screen |
Once the user gets over the initial shock, and reads the ‘Tip of the Day,’ which still pops up, they may start to explore the new features (as well as look for a way to produce the old, familiar JMP 8 screen!). The newest features are listed in Table 1. For even more detailed listings, go to www.jmp.com/software/jmp9/whyjmp9.shtml.
Other than its old staple of instantly producing excellent graphics for any analysis, this new version greatly extends the analytic capabilities with fast and easy connections to not only SAS, but Excel and R, two of the most useful programs for data analysis. Of further importance to some users is the ability to display data on geographic maps to do sophisticated demographic analysis.
As one who has a great deal of use for data mining techniques, the upcoming edition of JMP Pro will greatly expand the ability to generate robust predictive models far beyond those possible in earlier versions. The same can be said for JMP Genomics 5 in the area of pharmacogenomics, which will be reviewed here in the near future. Now, let’s delve a bit further into the excellent statistical/graphical capabilities…
We’ll use a simple medical chart of cholesterol values over time (Figure 2a).
![]() Figure 2a: Cholesterol drug data |
We do a quick distributional analysis to see the general data spread and look for aberrancies (Figure 2b).
![]() Figure 2b: Distribution of cholesterol values with time |
Other than a few low values, we don’t see anything too extraordinary. It appears that the cholesterol level is decreasing with time and, to confirm our suspicions, we can do a simple XY plot. With the Tables menu, we quickly subset the time and concentration data, stack them, and recode the times as a continuous, numeric function. We can then do the X by Y plot (Figure 2c).
![]() Figure 2c: X by Y plot |
JMP will automatically do the ANOVA for the fit, give the equation of the line and do a lack-of-fit test (among many other things). We note at this point that we have combined the A and B drug data and left out the control and placebo. To see if groups are similar, we can very quickly do both clustering (Figure 3a)
![]() Figure 3a: Hierarchical clustering |
and discriminant analysis (Figure 3b).
![]() Figure 3b: Linear discriminant analysis |
It is immediately apparent that Drug A is similar to Drug B, and the Placebo is similar to the Control. Also, the two pairs are dissimilar to each other. The cluster also colors the groups and lets us know that the data in row 6 (Drug B) is slightly different than the other data in that group. We had previously observed (in the Distribution Platform) that this was due to lower values in that patient than the others.
This was a simple example of data exploration, which is just the first step in an analysis. JMP 9 will do the usual, and more complex, model fitting and modeling, as well as testing the drugs’ quality attributes in the new degradation platform. Although scientific applications were used, JMP 9 has a really nice “Choice” tool for business applications and the Time-Series analysis tool for econometrics. I was particularly impressed with the way the new simulator features have been added to the very useful profiler to add noise and investigate variability for prediction studies.
All of the manuals are available electronically under the help menu, and are still easy to navigate (although I continually wish for upgrades to the indices and search function). Other helps available include Web sites, the excellent support team at the Help Desk, tutorials and regional/local users groups.
Although it is next to impossible to do justice to modern statistical software in short reviews, the readers are always encouraged to explore on their own at the company Web site and download the trial software.
I will end this review with a disclosure, as I have strived not to sound overly glowing in my praise. Your editor has founded two JMP user groups and regularly uses SAS and JMP (amongst others) in daily analyses.
Table 1: Key Features in JMP 9
New Platforms
The new Degradation platform lets you analyze product deterioration data over time to help you predict product quality and warranty risk.
The new Neural platform replaces the Neural Net platform and adds enhancements, such as richer diagnostics, including several measures of fit and diagnostic plots. With JMP Pro, choose which data to use for cross-validation. Also in JMP Pro, get automated handling of missing data and automated transformation of input variables.
Design of Experiments (DOE)
In Custom Design, choose a new optimality criterion: Minimum Aliasing Design.
In Custom Design, model and alias terms are available, and an alias model always appears.
In Custom Design, a color map of correlations of all model terms and aliasing terms appears.
Accelerated Life Test Design, a new DOE platform, lets you design high stress tests to simulate failure quickly — so you can find product weaknesses more easily and faster.
In demonstration plans, design a test to compare the reliability of a new product to a standard.
In reliability test plans, determine the sample size or length of study needed to obtain a given precision about a fitted quantile or probability.
This feature is available only in JMP Pro.
Interface to R
Interact with R using JMP Scripting Language (JSL).
Submit statements to R from within a JSL script.
Exchange data between JMP and R.
Display graphics produced by R.
Graph Builder
Incorporate geographic maps into graphs.
Graph two independent Y variables on separate axes.
Work with subsets of large data tables using the new Sampling function.
See the shape of your data using density contours.
Plot error bars and confidence intervals.
A new drop zone under the Group Y area supports a Frequency (or Weight) variable.
Graph Builder uses new custom color scale and gradient capabilities.
New Windows Environment
User interface for Microsoft Windows has been completely revamped.
All windows are now independent of one another, so it’s easier to use JMP on multiple monitors.
The Home window lists JMP files opened recently and all open JMP windows.
Partition
Use bootstrap forests to help ensure your predictions will generalize well.
Employ boosted trees to increase the accuracy of your predictions.
Use a cross-validation column to easily compare different model types and make sure that predictions generalize well.
Microsoft Excel Add-In
Use the JMP Profiler to visualize models contained in Excel spreadsheets.
Optimize and simulate using your Excel spreadsheets.
Use the add-in toolbar to copy data from Excel into JMP directly.
Availability
Single user $1,895 (single license, corporate)
Academic, student (call for pricing)
Site licenses available
SAS Institute
SAS Campus Drive
Cary, NC 27513-2414
1-877-594-6567
www.jmp.com/about/contact.shtml
John Wass is a statistician based in Chicago, IL. He may be reached at [email protected].