Definitive Screening Designs
Now you can design experiments to separate the vital few factors that have a substantial effect on a response from the trivial many that have negligible effects. If a factor’s effect is strongly curved, a traditional screening design may miss this effect and screen out the factor. And if there are two-factor interactions, standard screening designs with a similar number of runs will require follow-up experimentation to resolve the ambiguity. Wouldn’t it be nice if you could avoid both of these problems at once? With definitive screening designs, you can.
Excel Import Wizard
You know the situation – you’ve just received several Microsoft Excel workbooks that you must import into JMP for analysis. However, because the data is spread across multiple workbooks, there are nested hierarchies, and there are grouped rows or grouped columns, you need to do lots of work before you can start analyzing. With the power and flexibility of the new Excel Import Wizard, many of these data import pains disappear; you can now get to an analysis-ready JMP table in fewer steps, with less cleanup and reformatting.
- The Excel Import Wizard:
- Provides more options for data location during import.
- Aids verification of import settings by presenting a visual preview.
- Concatenates multiple worksheets and tries to match columns to ease data cleansing and joining tasks post-import.
- Provides stack capability to arrange data in a way that works better in JMP.
- Allows you to fill in values for Excel "merge" cells.
- Automatically generates JSL to refresh data from the Excel workbook, if the file changes or is updated.
Note: Excel Import Wizard is a Windows-only JMP feature.
Sometimes you need to transform variables, create ratios or express dates in different formats before you can move forward with an analysis. Now, JMP does this for you with a single click, letting you stay in the analysis flow – there is no need to stop what you’re doing to create a formula column or modify your data table.
- Transformed variables:
- Make it easier to stay in the analysis flow, while looking at data from different perspectives.
- Are temporary, formula-based columns that can be created on the fly without having to save them to the data table.
- Include a suite of functions to transform, combine or for complete flexibility apply a function to any of the variables.
- Also can transform dates, aggregating or recoding dates with ease.
- Are accessed by right-clicking on any launcher dialog.
- May be stored to the data table, if you want to create a permanent column.
Consumer and Market Research
You already collect information about how customers use a product or service or how satisfied they are with your offerings. The resulting insight lets you create better products and services, happier customers and more revenue for your organization. JMP now includes a full suite of tools for performing customer/consumer research. In the past, you may have had to use one product for consumer research work and JMP for design of experiments. Now you can do both types of analyses using a single product, for a more efficient use of your most precious resource: your time.
Tools for performing these statistical analyses are now located in one convenient place: the Consumer Research menu.
- Use the following analysis platforms for analyzing your data:
- Categorical (survey analysis)
- Factor analysis
- Choice (conjoint)
- Item analysis
Categorical analysis is made easy in JMP 11 and can support survey questions in multiple formats, allowing for both detailed and compact reporting. You can also analyze multiple response questions, where your survey includes questions for which respondents can choose more than one answer. There is even a drag-and-drop interface for building more complicated survey analysis structures. You can output the results in crosstab and multiple report tables, use share and frequency charts, view mean scores across responses, and perform tests and comparisons. And when you're finished, you can easily output the completed analysis tables into an Excel workbook.
Have you ever encountered one of the following scenarios: "I have 10,000 variables – what’s changed?" or "I have one response, but many possible predictors – what do I do?" If you have, you know that the large number of variables (or many levels of a categorical variable) can make it difficult to reach meaningful conclusions.
Before JMP 11, you had to run many analyses and look through potentially thousands of analyses or reports and if your data contained outliers, they could inflate the variance estimates and mask the significance of the effects.
In JMP 11, when data has many Y’s and X’s, selection bias or is messy, you can use the Response Screening platform to extract meaningful conclusions and tell at a glance what is important – all within an intuitive graphical report. With the robust regression (also known as a Huber regression/M-estimate) option, you can iteratively and automatically fit a model, protecting you against the influence of outliers while saving you from tedious data cleansing and manipulation.
- JMP 11 also contains many other features to help you handle messy data:
- Additional robust fitting options in Distribution, Bivariate and Oneway platforms.
- Fit Response Screening, a new personality in Fit Model, which lets you perform multifactor response screening, specify model structure and add interaction and polynomial terms to your models.
- Use the False Discovery Rate (FDR) report in the Response Screening platform to determine which effects are truly significant and which are just due to random variation. (FDR is a general way of doing multiple comparisons correction and helps to dramatically reduce false positives.)
- Option to create missing values as informative categories in the Partition platform.
Plotting your organization’s location-based data on geographic regions such as countries, states, or counties is an effective way to transform this data into useful information. It also allows you to quickly and visually spot trends and relationships that would otherwise remain hidden. Adding background maps to graphs has been built into JMP for several versions, but with JMP 11 you can plot your data on street-level maps, giving you access to geographic features such as cities, roads or bodies of water. These additional details give geospatial context to your data, providing you with additional insights that would otherwise be difficult or impossible to unearth. These detailed graphs can also be compelling communication tools when sharing your discoveries.
SAS servers host map data that create the images from open source maps available from OpenStreetMap (OSM). These servers can generate and return the maps when you select Street Map Service from any platform in JMP that supports background maps.
Assess Variable Importance
The Profiler provides a number of highly interactive cross-sectional views of any response surface. When your models have only a few effects, it is easy to see what the key drivers are. However, with big models, the task of find big effects through visual inspection can be tedious.
There are many reasons you’d want to be able to effectively assess variable importance in your models: to better see and understand the most important inputs of the phenomenon being modeled, to know how changing certain factors may affect the outcomes, and to determine which factors or combination of factors might be influenced to create better outcomes. JMP now lets you perform this assessment in a single click with the Assess Variable Importance routine in any Profiler. Using this tool also provides a common method for assessing variable importance across multiple modeling methods, which may have varying ways of assessing the goodness of fit.
Choose from the following sampling methods of the input variables to assess variable importance:
- Independent uniform inputs – a simple drawing of points randomly across the range of the input variables.
- Independent resampled inputs – a bootstrap sample that assumes independence among the input variables.
- Dependent resampled inputs – a bootstrap sample, which draws rows while taking into account the correlation among the inputs.
A summary report lets you perform sensitivity analysis and shows each column’s main effect and total effect importance. From the variable importance report, easily reorder the Profiler by main or total effect importance, or colorize the Profiler by effect importance. This is a huge time-saver when models may have dozens of predictors, letting you separate the vital few factors that may be driving the response. You may also want to use this technique for variable selection, utilizing the driving factors to fit additional, more parsimonious models.
Smarter Filtering and Summaries
Statistical discovery requires great flexibility in how you view, filter and work with your data. With the powerful tools in JMP 11, you’ll gain more insight, in less time.
The Data Filter and Local Data Filter make filtering more efficient, with options including delimited multiple responses, "find," "contains," inversions, conditionals, favorites and more.
A new tool for summarizing data table structures, the Columns Viewer, will quickly become one of your favorite new features, especially if you have complex, wide data tables or work with data tables created by others. In one click, find columns that contain formulas, have spec limits or any multitude of other column properties – and quickly access column information to view or change any column parameters.
During exploratory data analysis, the quick subsetting and summary statistic capabilities of the Columns Viewer are invaluable. Select a number of columns, access summary statistics and launch the Distribution platform right from the viewer to spot patterns in your data without having to navigate through potentially hundreds or thousands of columns.
Also, a more efficient table sort and general improvements to the Column Switcher let you search by column type, name and other shared attributes. No matter how you work with your data, JMP 11 lets you do it more efficiently.
Save Interactive HTML Reports with Data
Interactive HTML enables JMP users to share dynamic graphs and reports, so that even those who don’t have JMP can explore the data. The JMP report is saved as a Web page in HTML5 format, which can be e-mailed to users or published on a website. Users then explore the data as they would in JMP. The report can be viewed in most modern browsers, including those on mobile devices.
- Interactive HTML provides a subset of features from JMP:
- Explore interactive graph features.
- View data by brushing.
- Show or hide report sections.
Many customizations to the JMP graphs, such as ordered variables, horizontal histograms, background colors, and colored data points, are saved in the HTML file. Graphs and tables that were closed when the page was created remain closed in the Web page until opened. JMP creates interactive HTML for features in many platforms.
When you save reports as interactive HTML in JMP, your data is embedded in the HTML. The content is unencrypted, because web browsers cannot read encrypted data.
Build Attribute and Rare Event Charts
The drag-and-drop Control Chart Builder is an easy way to create control charts of your process data with a single click. Like Graph Builder, you select the variables (or columns) that you want to chart, and then drag and drop them into zones. The instant feedback encourages further exploration of the data. You can easily fine-tune the visual display on the current chart, or, if you change your mind, quickly create another type of chart. The Control Chart Builder now lets you build attribute charts, including the np-, p-, c- and u-charts, which are helpful to quality practitioners. Because some events, such as infections in a hospital, occur so infrequently that a traditional chart is inappropriate, you can also build g-charts. The g-chart is an effective way to understand whether rare events are occurring more frequently than expected and warrant an intervention. Or build rare-event t-charts, which are based on the Weibull distribution and are used to measure the time that has elapsed since the last event.
An Event Chooser is also provided for real-time exploration of attribute charts. This enables you to select the event or events of interest for the attribute chart, updating the p- or np-chart automatically as your selections change.
Crisper, Cleaner Design
With JMP 11, you get better default graphics and tables: extensive new preferences that let you fine-tune the appearance of JMP graphs, tables and reports; and graphs that apply best practices for data visualization. These allow you to create effective visualizations automatically, minimizing the need to adjust the output of JMP, all the while placing more emphasis on your data and its message.
- New features include:
- Updated Graphs Preference Group, which lets you refine how JMP Graphs appear in the analysis platforms.
- New Styles Preference Group, which lets you adjust how tabular reports appear throughout JMP. Preview the adjustment just as you would in the Graphs Preference Group.
- Statistical effects that are color-coded by their significance level.
- More informative labels and hover-help, with support for pinning multiple labels to a graph.
- "Anti-collision" smart labels, which increase readability and emphasize outliers in graphs.
- New color themes and more efficient creation of custom color themes. Themes are separated into sequential, diverging, qualitative and chromatic themes so that you can pick the best color scheme for the data being graphed.
Import Sampling for SAS® Data Sets and Text Files
Often a SAS data set is too large to open up entirely in memory and on your desktop machine. Or, for exploratory data analysis, it may be advisable to sample a subset of the data set for more efficient modeling, visualization and analysis. The same may be true if your data resides in large, flat text files. Before JMP 11, it was difficult to deal with these enormous data sources. Even when taking a sample of the data, the entire data set first had to fit in memory. JMP 11 now uses powerful routines from SAS’ data mining software, SAS® Enterprise Miner™, so you can sample data sets before importing them into JMP. You have complete flexibility in how you take the sample, so fewer steps are required for pre-analytic data cleanup and preparation. The next time you are faced with a massive SAS data set, remember that you can conduct exploratory data analysis, modeling and visualization on a subset of the data. Similar sampling provisions also exist for text files that would be otherwise too large to fit in memory.
Engineers who have a large investment in custom MATLAB models, programs or algorithms can now interface directly to MATLAB from JMP using new JSL functions in JMP 11. Initiate a MATLAB connection, send data to MATLAB, submit code, and bring data or output back to JMP. Or use the Application Builder in JMP to build customized GUIs, which run simulation models in MATLAB and return the results in JMP for further analysis. With JMP 11, you can enable others to use your MATLAB models even if they know nothing about MATLAB. To extend the functionality of JMP even further, you can use MATLAB’s external programming language interfaces to employ functionality from other languages in MATLAB, and then return the results to JMP.
- Other options for JMP and MATLAB interactions include:
- Send a JMP data table to MATLAB to use engineering modeling techniques, and then return the results to JMP for visualization and profiling.
- Use the Custom Designer in JMP to optimize complex MATLAB models: generate experimental runs, execute a MATLAB model and return results to JMP for analysis.
JSL Development Environment
Build custom applications more easily with a host of new features in Application Builder. These features include:
- More flexible window combinations, including stretchable window arrangements.
- Support for embedding data tables within applications.
- Multiple data table support.
JMP Scripting Language (JSL) users can now develop, debug and deploy scripts more intuitively, making custom application creation more efficient. Improvements to the JSL development environment include:
- A JSL Profiler, which analyzes scripts to find code bottlenecks.
- An improved Show Tree Structure interface that helps identify specific JMP report elements.
- New scripting options for Graph Builder, Column Groups, DOE and triangulation functions.
- Updated scripting index with the latest JMP 11 functions, including sample scripts.
- Improved error logging and annotation – the JSL log now highlights the line of code that is failing in your scripts and provides more informative error messages.
- Code Folding to expand and collapse marked blocks of JSL code.
Multiple Comparisons Made Easy
When comparing more than two means, an ANOVA F test tells you whether any of the means are significantly different from each other, but it does not tell you which means differ. Multiple-comparison procedures give you more detailed information about the differences among the means – and with JMP 11, it is easy to perform multiple comparisons in a variety of contexts. The goal in multiple comparisons is to compare the average effects of three or more "treatments" (for example, drugs or groups of subjects) to decide which treatments are better, which ones are worse, and by how much – while at the same time controlling the probability of making an incorrect decision.
- Multiple comparisons can:
- Be added as a parameter estimates model option in the Fit Model report.
- Involve a single effect or be defined for flexible custom comparisons.
- Compare effects to the overall mean, or obtain all pairwise comparisons using Tukey’s HSD or Student’s t.
- Be communicated to, and understood by, others more easily by presenting graphical output along with the statistics.
- Use least square means estimates or user-defined estimates.
- Also conduct equivalence tests.
JMP® 11 New Features Overview
Design of Experiments
- Definitive Screening Design platform identifies crucial factors affecting a response.
- Flexible orthogonal and near orthogonal array generation supports factors at mixed levels and arbitrary numbers of runs.
- Cluster-based space-filling designs allow for inequality constraints on factors.
- Space-filling mixture designs are now supported.
- Many DOE platforms are now scriptable.
- Revamped and enhanced power analysis section of the design evaluation output in most DOE platforms.
Quality Engineering, Reliability and Six Sigma
- Attribute Charts (np-, p-, c-, and u-chart) are supported within the Control Chart Builder.
- Event Chooser in Control Chart Builder for real-time exploration of attribute charts.
- Rare-event g- and t-charts are supported within the Control Chart Builder.
- Work with data sets with one-sided specification limits in the Capability platform.
- Support for Wheeler’s range-based measurement in Measurement Systems Analysis platform. Nested and crossed model types available for EMP MSA method.
- Ability to change the process sigma value in the Shift Detection Profiler in the Measurement Systems Analysis platform.
- Bayesian estimation of life distribution fits now supports Quantile and Parameter priors as well as Failure Probability priors in addition to the current Location and Scale priors.
Statistics, Predictive Modeling and Data Mining
- Response Screening platform extracts meaningful conclusions from large, messy data sets.
- Robust fitting options in many platforms.
- "Informative missing" option to treat missing values as informative categories.
- Least Squares personality in Fit Model supports multiple comparisons of Least Squares means as well as many custom estimates.
- Assess variable importance, an omnibus method for assessing variable importance across multiple modeling methods and accessed through a Profiler. This feature allows you to understand factors most affecting outcomes they are modeling.
- Profit matrix specification when want to convert a prediction model to a decision model by assigning weights to outcomes.
Data Visualization and Exploratory Data Analysis
- Graph Builder offers a multitude of improvements, including the ability to create geospatial scatterplots, to page through group variables and to add more summary statistics to graphs.
- Windows-only Excel Import Wizard provides a preview of the data so that you can modify its structure before the import.
- Transform Columns creates virtual formula columns within a platform or launch dialog, or through JSL.
- Save interactive HTML output to share reports that contain dynamic graphs.
- Plot data on street level maps, with access to geographic features such as cities, roads or bodies of water.
- A Columns Viewer finds important columns and generates summary statistics.
- SAS® integration and text import tools in JMP provide new sampling options.
- Data table compression reduces file size by up to 90%, making it easier to transfer and share data tables.
- Multiple JMP tables and reports can be combined into a single window on Mac and Windows, with more flexible options for layout hierarchies and resizable content.
- Tabulate is now a platform in the Analyze menu. Improvements include removal of the popup option for role selection.
- New labels and automatic label placement options make interesting data points easier to locate and identify.
- Extensive new preferences help fine-tune the appearance of JMP graphs and reports.
- New options for customizing color themes.
- Expanded column filtering options search by type, name and other shared attributes.
- Data Filter now saves favorites and easily negates whole sets of filter conditions.
Consumer and Market Research
- Methods for understanding consumer preferences have a new home in the Consumer Research menu. The reorganized menu addresses the needs of behavioral/market research users or people analyzing survey data.
- Menu includes Categorical, Factor Analysis, Choice and Item Analysis.
- Categorical platform improves response roles, report formats, and offers drag-and-drop creation of structured responses.
- Ability to import data from IBM SPSS Data Collection Survey Reporter.
- Leverage Excel import wizard if survey data is stored in Excel spreadsheets.
- Data filters let you build hierarchies, save favorite data filters and build summary categories.
- Factor Analysis platform supports multiple fits and rotation, conditional formatting to suppress small values, and provides commands to sort and save factor patterns.
- The Choice platform includes new options to perform comparisons between specific alternative choice profiles and calculate price changes equivalent to feature changes.
Analytical Application Development
- Application Builder now offers more flexible window combinations, including stretchable window arrangements for embedded data tables and resizable graphics.
- Multiple data tables support has been added for JMP Applications.
- Connect to MATLAB with JSL scripts.
- The JSL Profiler helps to optimize scripts.
- An improved Show Tree Structure interface helps identify specific JMP report elements.
- New scripting options for Graph Builder, Column Groups, DOE and triangulation functions.
- Log with error reporting.
- Support for SAS 9.4 IOM interface.
- Code Folding to expand and collapse marked blocks of JSL code.
- Splitter Box enables you to create display layouts that can be dynamically resized.
Download a PDF of the new features in JMP and JMP Pro or view our online searchable documentation.