JMP 8: Continuous Improvement
The latest version of this statistical discovery software offers stunning graphics quality
click to enlarge
Figure 1: The Graph Builder
It is always a pleasure to greet an old friend and, with each new release of JMP statistical/graphics software from SAS, excitement never abates. This major upgrade includes additions to the statistics, graphics, QA, ease-of-use, SAS integration and data management areas. In fact, the ‘New Features in JMP 8’ listing runs to 19 pages. A “brief” overview is presented in Table 1 and operating system requirements in Table 2.
Keep in mind that many of the best features of the new software are dynamic visualizations of the data (examples include by time and ID variables such as country, percentage of population, etcetera) that are constructed with the newer bubble plots. The scatterplots with nonparametric density contours and marginal distributions are literally breathtaking and will display remarkable amounts of complex data in a form highly amenable to pattern discovery. For scientists and engineers searching for understanding of multivariate processes, these new graphics provide the tools for breakthrough discovery.
click to enlarge
Figure 2: Bubble plot with tails
To introduce but a small array of the newer features present, let’s concentrate for a moment on the really nice application in software development called the Graph Builder. This revolutionary feature allows a series of drag-and-drop operations to build new graphics with ease and maximal flexibility. For those who used the ‘Tabulate’ function in version 7, this will be an even easier natural extension. By using the .jmp file “Big Class,” dragging and dropping the height and weight functions to the graph area and sex to the ‘overlay’ area, Figure 1 was constructed in a few seconds. It quickly shows us that weight will increase with height and that males tend to be larger than females (a really electrifying discovery).
The information above pales in comparison to those that we may extract from the new bubble plots with tails (Figure 2), the nonparametric density plots with marginal distributions (Figure 3) and the newer profilers (Figure 4).
Lest the new user be overwhelmed with the quality of the graphics, we need to briefly mention the newer analytic features. JMP 8 has a new unit designed particularly for business decision-driven survey analysis. Using the new ‘Choice’ platform within the Analyze/Modeling menu, the analyst will make use of the Fit Model and Prediction Profilers already in earlier models of JMP to determine which settings of the variables (choices presented to potential/actual customers) are most desired. This will, if used judiciously, maximize profits. A simulator has been added to increase the accuracy of the estimates and to provide greater predictive power.
click to enlarge
Figure 3: Nonparametric density with marginal distributions
For those into reliability analysis, the new ‘Fit Life by X’ platform will perform some rather amazing analytic feats with just a few clicks on the mouse (assuming a knowledgeable analyst). By specifying a time-to-failure, predictor variable, any weights and censoring factors, as well as a proposed distribution, a number of diagnostics are produced. These allow failure analysis with accelerating factors under a wide variety of assumptions. The addition of Flash Output capabilities now allows sharing the graphics (and embedding them in documents) with colleagues who do not have JMP.
Myself, I’m an ardent fan of the Analyze menu, but especially the Distribution and Fit Model Platform. Under distribution, things can get a lot more sophisticated than you would think. The new GLog (Generalized Logarithm) is a transformation to the normal distribution and will greatly aid scientists with many types of assay data, including those that violate the constant variance requirement. To assist the analyst with a quick check, the new ‘All’ choice will fit all of the continuous distributions and issue a ‘Compare Distributions’ report which will sort them in order of the adjusted Akaike Information Criterion (AICc) and automatically fit the distribution with the lowest AIC. In this case, the developers saw fit to adjust the AIC by the number of parameters and sample size. In most cases, I would recommend against using a single criterion, but here the adjustment gives us more confidence in the diagnostic. There are a number of new diagnostic tests for model fitting and the reader should check the Web page for a full listing.
click to enlarge
Figure 4: Prediction profiler with simulation/optimization
As with most modern statistical and mathematical software, the developers have strived for simplicity and a gentle learning curve. It is no longer necessary to have advanced degrees in statistics to use these programs, and the novice can be up and running in JMP in a matter of minutes, and somewhat proficient in a few hours. Of course, the more that one is versed in statistics, the better.
To assist in flattening the curve, JMP provides many helps. There are excellent written manuals and books by users, electronic searchable manuals within the program, on-line FAQs, an excellent support desk and numerous regional users’ groups. (Disclosure: Your editor is a founder and active member of two such groups.)
This latest version is stunning in the quality of the graphics, and JMP has pioneered the advancement of statistical graphics by heavily linking most number crunching operations to a graphic. Interested parties are highly encouraged to download a trial version.
• $1,495 single user (corporate) • academic, student and site licenses available
JMP, a business division of SAS
SAS Campus Dr., Building S
Cary, NC 27513-2414v 1-877-594-6567; 1-919-677-8000
John Wass is a statistician based in Chicago, IL. He may be reached at editor@ScientificComputing.com.
Table 1: Some Key Features in JMP 8
• Graph Builder platform lets you construct graphs interactively as you explore your continuous and categorical data.
* Drag and drop to instantly create scatterplots, bar charts, histograms, trend lines, box plots and more.
* Visually compare across grouping levels for insightful analysis.
* Partition data by both continuous and categorical variables to easily see interactions.
• Choice platform helps you discover consumer preferences and tailor products and services features to consumer desires.
* Discover differences in market segments to create targeted offerings.
* Use what you know about consumer preferences to design your study.
* The integrated Profiler helps you understand, visualize and optimize products and services.
* Get utility scores for segmenting/clustering your data.
* Use Firth bias-corrected maximum likelihood estimators.
• New additions to the Reliability platform improve your understanding of product and part lifetimes.
• Use Fit Life by X to analyze lifetime events for accelerated test data. Compare different distributions at the same factor level or the same distribution across different factor levels.
• Use Life Distribution to find the most suitable distributional model for your time-to-event data and make predictions. Weibull, Lognormal, Fréchet, Extreme Value and other distributions are available.
• Use Competing Causes analysis to explore failure causes in a system and understand the effects of repair.
• Mixture Profiler shows response contours for mixture experiment models, where three or more factors in the experiment are components in a mixture.
* Visualize and optimize response surfaces resulting from mixture experiments.
Integration with SAS
• JMP 8 works with SAS 9.1.3 and SAS 9.2.
• A new SAS toolbar provides convenient access to SAS features.
• See and import data from your local SAS server connection in the Browse SAS Data window and through JSL.
• Take stratified random samples (using PROC SURVEYSELECT) when importing data from SAS.
• Get SAS results in HTML, PDF or RTF when you submit SAS code.
• New data sets created by submitted SAS code are now tracked and imported into JMP automatically.
• Get SAS procedure output in the familiar JMP report format.
• Export Bubble Plots and Profilers as Adobe Flash animations for use in Web pages, presentations and other documents.
Languages and Currency
• JMP 8 handles international character sets with ease.
• Use JMP interfaces in English, Japanese, Chinese, French, German and Korean.
• Include multiple currency types in a single data table.
• Use new Bubble Plot Trail Lines to connect bubble trails for enhanced visualization of change over time.
• Contour Plot now has a Color Theme menu.
• All Control Charts can now shade zones.
• Surface Plot interface has been simplified.
Design of Experiments
• Design optimal Choice experiments for analysis in the Choice platform.
* Use the new design diagnostics outline node for all JMP designs.
* Automatic recalculation
• Auto Recalc is now available in more platforms.
Variability and Gauge Charts
• A frequency column option is now available in the Variability launch window.
• The Variability platform has a new way to compute variance components using Bayesian estimation.
• A new bias report and graph shows the bias by each standard reference value used.
• Misclassification probabilities are now available for the Gauge R&R studies.
• Specify limits in Gauge R&R.
• Get parameter-estimates-type reports in the linearity report.
* Test both the slope and intercept.
* Test the slope to look for linearity issues.
• Make variability and standard deviation graphs vertical.
• Goal plots can now draw a defect rate contour and shade Cpk regions.
• Interactively change the displayed Cpk threshold using a slider bar.
• Points appearing in goal plots now link directly to their respective columns in the original data table.
Table 2: System Requirements
• JMP/SAS Integration: SAS 9.1.3 SP4 or SAS 9.2 required
• OS: Windows XP w/ SP2, Windows Vista (Vista Home Basic edition not supported)
• CPU: Pentium 4, Athlon and higher
• RAM: 512 MB minimum (1G+ recommended)
• Drive Space: 130 MB free disk space (200 MB recommended)
• Browser: MS Internet Explorer 6.0 or higher
• Recommended Display Settings: True (24 bit+) color with resolution of 1024x768 or greater; video card with hardware accelerated 2-D and 3-D drivers recommended
• Database: UNICODE compliant ODBC 3.0 or higher (required only if connecting to database)