New in JMP 13
General Usability Enhancements
- Reorganized menus incorporate new capabilities into Analyze and DOE menus.
- New JMP Starter and JMP Pro Starter identify differences between JMP and JMP Pro.
- Green run button streamlines running scripts from Table window.
- Windows only: Quick Open makes files easy to search and find.
- Friendlier Formula Editor provides full-screen editing, formula term search and filtering.
Getting Data into JMP
- Conditional filters at import runtime help shape data when joining or updating joined tables.
- Queries save to data table for reuse, refresh and update using new data.
- Virtual joins combine tables in temporary memory when you want to keep tables separate in order to distribute table maintenance and minimize RAM requirements.
- Column Panel lists virtual join table and column information.
Streamlining Design Comparison
- One report compares statistics and graphical views of correlations, prediction variances, aliasing and efficiency for up to three designs with unequal number of runs.
- Report automatically updates when you add or remove model terms or when expected coefficient estimates or standard errors change.
Enhancing Data Exploration using Graph Builder
- Parallel Plot handles interactive interrogation of continuous or categorical data.
- Tree Maps can incorporate any number of nested values. New interactive menus let you configure maps to your specification.
- Bar graphs and position legends can be displayed inside graphs and axes can be customized to many level of complexity.
- Four new methods graph missing data.
- Different graphic transformations are available from Graph Builder and other platforms.
Using Unstructured Text
- Text analysis techniques enable using relevant text variables in data exploration and analyses.
- Custom rules let you specify guidelines for processing text.
- Term and phrase lists, recode from within Text Explorer, key words and phrases examination by document (row), and stemming identify patterns in text.
- Indicator values (flags) identify rows containing a particular word or phrase.
- Word clouds shaded interactively by variables you select help identify text patterns.
- Graph Builder summarizes, analyzes and visualizes prevalence of identified words or phrases in your data.
Screening Multiple Processes
- Process Graph reports overall process health.
- Customizable tests and metrics screen large number of processes to quickly identify processes of interest that warrant further analysis.
- Interface makes it easy to search, sort and graph by metrics and interactively select one or more tests for special causes (Western Electric rules).
- New Process Screening Shift Graph and Process Performance Graph compare processes.
Analyzing Non-Normal Process Capability
- Process capability reports run from one platform, regardless of process distribution normality.
- New Capability Index Plot displays values for all variables, including variables modeled with non-normal distributions.
Building Dashboards and Using New Ways to Save Results
- Dashboards compile reports that can be drilled into, filtered and refreshed based on filters.
- When Dashboard is re-rendered, data for data tables built from query automatically updates.
- Template and drag and drop builds Dashboards that can also be reconfigured using drag and drag.
- Graph Builder generates fully interactive HTML reports for all element except Contour Plots, Pie Charts, Parallel Plots, Tree Maps and Formulas.
- Customizable consolidated Web Report thumbnail display provides an interface to interactive reports.
- Multiple-file export saves more than one JMP file to Excel Workbook.
New in JMP Pro 13
Selecting and Organizing Unstructured Text for Analysis and Modeling
- Choose which JMP Pro feature to use for different text exploration tasks.
- Identify text terms that tend to appear together.
- Group and explore similar documents using clustering and Latent Class Analysis.
- Report recurring themes and documents.
- Use text data to build predictive models.
Storing and Deploying Prediction Models using Formula Depot
- Store prediction formulas and model metadata in one place (Formula Depot)
to access easily when comparing models.
- Add Profilers and Comparison Reports to Formula Depot.
- Generate score code in C, SAS or SQL for sharing models outside JMP.
Using Generalized Regression for Variable Selection, Linear Modeling and Analyzing DOEs
- Handle left-, right- and interval censoring to enable fitting time-to-event models with variable selection.
- Use penalized methods for no-intercept models.
- Use Cox Proportional Hazard models.
- Use new ERIC fit statistic that is generalization of BIC especially for the LASSO and deploy new double-LASSO to increase prediction accuracy.
- Employ two-stage Forward Selection for OLS-based modeling when potential effects outnumber rows.
- Examine interactive ROC and lift curves for categorical responses, and residuals by predictor plots for continuous responses.
- Build simulations that use parametric bootstrapping and power estimation in nontraditional settings. or when expected coefficient estimates or standard errors change.
Using Association Analysis to Identify Similar or Related Items
- Use Transaction Listing to make tall datasets wide.
- Convert character column with delimited fields to multiple columns.
- Use Frequent Item Sets to see percentage of items with a particular word value.
- Interpret association rules, rules confidence levels and lift values for similar or related items.
Simulating Repairable Systems
- Describe how expensive, repairable systems behave and forecast expected component or system failure times.
- Predict service time and system availability during different time periods.
- Determine which components or systems should be simultaneously maintained or repaired.
- Forecast cumulative breakdowns and repair costs over a given time horizon.
Building Complex Mixed Models
- Use new covariant structures including Unequal Variances, Exchangeable (Compound) Symmetry, Exchangeable Unequal Variances, Toeplitz with or without unequal variances, and Antedependent with or without unequal variances
Using Partition to Build Complex Decision Trees Faster
- Specify random seed value and suppress multi-threading to ensure repeatable results for Bootstrap Forests and Boosted Trees.
- Use tuning design tables when deploying Bootstrap Forests and Boosted Trees.
- Fit boosted trees faster using Stochastic Gradient Boosting that also makes models more robust to overfitting.
Using Naive Bayes for Categorical Responses
- Fit models to predict the value of a categorical variable using fast and computationally simple Naive Bayes method that is especially suitable when there are a large number of predictors.
- Save prediction formulas for model comparison.
