INTERNAL USE ONLY: This document is internal use only at this time.
This is a feature article for the JMP 13 Foreward about JMP 13.
(JMP Foreword Article) – Daniel Valente 5/11/2016
INTRODUCING JMP 13
Is there gold in your data?
JMP 13 helps you make your own luck; discover how new analysis platforms, feature enhancements, and improvements across the entire analytic workflow—from importing data to sharing analysis results—help you find the unexpected in your data.
Here are the new things in JMP 13 that you will want to explore:
1) Query Builder for JMP tables
The Query Builder for relational databases and SAS data sets has quickly become the preferred way to easily create accurate, repeatable and sharable queries—without having to write SQL code. However, until now you could not use Query Builder to join JMP tables already in memory, and nobody wants to create a database just to join data that is already in JMP. JMP 13’s Query Builder for JMP tables provides multi-table query and join capability through the same interface as the SQL Query Builder and SAS Query Builder.
Once you start using JMP 13’s Query Builder, you’ll wonder how you ever lived without it.
In addition, JMP 13 has many improvements that serve to streamline the data preparation workflow. Here are just a few:
· Run a script in a data table with a play button.
· Right-click “new column” options in the data table.
· Create and view formulas more easily with a retooled formula editor.
· Access the Local Data Filter in the main red triangle menu, rather than in the script submenu.
· Use SQL as a query language in JSL.
2) Dashboard Builder
Often, scientists and engineers must sell the change solution to others who may lack their level of analytic expertise. For clear and efficient communication, a single presentation-ready dashboard is often the gold standard. In previous versions of JMP, the process of creating a dashboard required many clicks. In JMP 13, the Dashboard Builder aggregates JMP reports into a presentation-ready dashboard with only a few clicks and includes such time-savings features as:
· Drag-and-drop templates.
· Bigger, more visible drop zones, like Graph Builder.
· One-click assignment of graphs as filtering devices.
· Real-time, customized analysis and layout reorganization.
· Tabbed dashboards made on the fly.
· An interactive HTML report that preserves the layout of your dashboard and can be shared with others who don’t use JMP or have access to it.
3) Virtual Join
A virtual join eases the pain of joining and/or maintaining large (or many) tables. In previous versions of JMP, tables had to be physically joined. In some cases, this required more memory than could comfortably be spared. Consider the scenario of joining a tall table to a wide one (for example, joining weekly summary values to data collected every second). Even in the best cases, the memory demand was high. In the worst cases, the join was impossible. With the virtual join in JMP 13 you get the benefits of a join, without creating a new table or expending memory. The resulting memory savings can be significant – both virtually and physically – especially when columns contain large strings or images.
4) Process Screening
In previous versions of JMP, there was no way to screen hundreds—let alone thousands—of processes automatically. As a result, process control analysts, Six-Sigma practitioners, semiconductor engineers and other quality engineers had to manually triage these reports, which meant that most of the time allotted to searching for “problem” processes was consumed in viewing processes that were well under control. In JMP 13, there is a better way.
JMP 13’s Process Screening platform lets you quickly determine overall process health with the Process Performance Graph, and easily identify, using a customizable combination of tests and metrics, which processes warrant extra attention. It also lets you easily create control charts and process capability reports for any processes you choose.
With all the time you will save, you can focus more on problem solving, and less on fire-fighting.
The Process Screening platform in JMP 13 joins Predictor Screening (formally known as Screen Predictors) and Response Screening, as well as other modeling utilities (such as Explore outliers and Explore missing values) in the new Screening submenu of the Analyze menu.
5) Text Explorer
Free text data exists in many forms: survey responses, repair logs, engineering reports and free-response fields are just a few examples. In previous versions of JMP, market researchers, warranty engineers, medical professionals, engineers and scientists who had mountains of text-based data could do very little with it in JMP besides counting and recoding words. JMP 13’s Text Explorer uses natural language processing (NLP) algorithms to search a collection of documents (called a corpus), in order to identify facts and relations in unstructured text data. With Text Explorer, you can easily triage and uncover the meaning in your text data, rather than having to choose to either process it manually, or ignore it altogether.
Text Explorer in JMP:
· Provides methods for basic keyword extraction.
· Has a local recode to clean up documents without altering the original text.
· Lets you find latent information in text data.
· Lets you process text data in informative ways.
6) Export Improvements
Anyone who shares JMP results outside of JMP can do so more easily and effectively in JMP 13.
In the past, it was time-consuming to build a summary website of JMP results—but without this, an individual report could lack context. Interactive HTML reports are useful, but a single report usually cannot capture the entire analytic process. JMP’s HTML Auto-Reports solved this problem by providing a linked, thumbnail-indexed collection of HTML reports—but installation of an add-in was required. In JMP 13, the HTML Auto-Report feature is built-in.
Also, the tedious and error-prone work of copying and pasting multiple JMP tables into Excel workbooks is now a thing of the past: create, update or append to an Excel workbook of multiple JMP tables—each in its own worksheet—with a single click.
7) Graph Builder Additions
One of the most loved tools in JMP, Graph Builder brings value to nearly every user, helping them make beautiful graphs to communicate their data’s story as effortlessly as possible. In JMP 13, a host of new improvements and additions appear:
· Parallel Plots are now supported.
· Treemaps support an arbitrary number of nesting levels, with much greater control over the labels.
· Alpha level can be specified for confidence bands.
· Ordering points is easier – just right-click the x-axis.
· Line element supports multiple “connect through” options and a centered step option.
· Legends, which have a new settings menu, can be placed inside the graph.
· Geometric mean has been added as a Summary option.
· The correlation coefficient can be shown for the ellipse element.
· The Response Axis property is now supported by all feasible elements.
· Multiple size and color variables are allowed.
There are also some notable new features that make life easier not only in Graph Builder, but throughout JMP:
· Quantile, Day of Week, Random Sampling and Moving Average one-click transforms.
· Formula-based custom formats for numbers.
· New color themes and lightness controls.
8) Interactive HTML Reports
We asked users which JMP reports should be available as interactive HTML, and the answer was nearly unanimous: Graph Builder! Message received. Now interactive HTML exports support most of the Graph Builder elements. We love granting wishes.
You’ll get interactive versions of Points, Smoothers, Ellipses, Lines, Bars, Areas, Box Plots, Histograms, Heatmaps, Mosaic Plots, Caption Boxes, and Map Shapes. We’ve also worked to ensure your eye-catching dashboards built with the Dashboard Builder remain just as attractive when they are rendered as interactive HTML.
9) Compare Designs
In previous versions of JMP it was difficult for DOE practitioners to compare designs easily, especially when there were more than two designs – it was a manual process requiring copy/paste or shuffling views among multiple design evaluations. Now, compare up to three competing designs and get all of the diagnostics in one report. Compare the designs with respect to power, correlation, prediction variance, aliasing and efficiency.
10) General DOE Improvements
With each new version, JMP improves upon its world-class design of experiments capabilities. In JMP 13, key improvements include:
· Support for Extra Runs within Definitive Screening Designs (DSDs). These designs maintain the attractive aliasing properties of a DSD, while offering increased power due to the extra runs. Options for zero, four or eight extra runs are available.
· A Fit Definitive Screening Design platform, which is a specifically tailored modeling technique for DSDs, has been added to the DOE menu.
· Simulate responses, which allows you to simulate data from a Normal, Binomial or Poisson distribution, as well as adjust the simulation parameters after the design has been created by running the DOE Simulate script in the data table.
· Updates to the DOE menu structure group the designers under descriptive headings, providing more intuitive access.
11) Non-Normal Capability Analysis
In previous versions of JMP, quality assurance analysts, quality engineers, supplier quality engineers and Six Sigma professionals could perform capability analysis for normally-distributed processes in the process capability platform, but for non-normal processes, had to turn to the Distribution platform. Also, JMP provided standard capability indices, but these only apply when the process’ distribution is normal. In JMP 13, Capability reports can be created in a single platform, regardless of the processes’ distributions. You can also:
· Independently set the distribution for each process in the report, and save these settings as column properties.
· Use nonparametric capability indices when the process distribution is non-normal.
12) Cumulative Damage Models
Design, manufacturing and test engineers need to be able to analyze time-to-event data under non-constant stress situations, but classical Accelerated Life Tests (ALTs) assume constant stress over the life of any given unit. JMP 13’s Cumulative Damage platform:
· Provides an alternative to classical ALTs.
· Supports analysis of accelerated tests that specify varying stress profiles.
· Supports analysis of field data under arbitrary stress scenarios.
· Analyzes data having Step Stress, Ramp Stress, Sinusoid or Piecewise Ramp Stress profiles.
13) Multiple Systems Reliability Growth
Before, engineers analyzing reliability growth on concurrent or parallel systems often had to use competing software. Now, JMP’s Reliability Growth platform offers the following new capabilities:
· Concurrent system analysis: Several similar systems are under study together. When any system experiences failure, the design correction applied to it is also applied to each of the other systems under study.
· Parallel system analysis: Several systems, which may or may not be identical, undergo testing in phases. A primary goal is to compare the reliability growth across systems.
14) Latent Class Analysis
Market researchers and social scientists need tools to reduce complexity in their data. Latent Class Analysis (LCA) is a statistical method for identifying unobserved class membership among subjects, using categorical observed variables. For example, an insurer may wish to categorize people into different driver categories (latent classes), based on the type of cars they drive, their accident records, and other publicly available data. JMP 13 adds LCA to the newly reorganized cluster submenu, which includes a suite of techniques: Hierarchical, K Means, Normal Mixtures and Cluster Variables.
15) Multidimensional Scaling (MDS)
MDS is a popular multivariate statistical technique used by sensory analysts, social scientists, market researchers and biologists to create a visual representation of the similarities among a set of objects. For example, given a matrix of perceived similarities among various brands of cars, MDS produces a map where points corresponding to similar cars are closer together and points corresponding to dissimilar cars are further apart. There is little formal inference, such as hypothesis testing, associated with MDS, but it is a good tool for comparing products and making judgments about similarities without having to list product attributes, as other techniques, like conjoint analysis and factor analysis, require.
16) MaxDiff
MaxDiff analysis is a technique used to analyze customer preference. Similar to choice analysis, respondents simply select the best and worst choice from a set of fixed options. MaxDiff studies are often easier to execute than standard choice experiments, and in certain contexts, they can be more informative. With JMP 13, researchers can not only analyze MaxDiff studies, they can design them using the DOE platform.
17) No-Choice Option in Choice Analysis
Many market researchers conduct choice experiments where “no preference” may be selected, instead of any of the other options. Ignoring this possibility, where it exists, produces biased analysis results. JMP 13 now supports this important option.
18) JSL Improvements
For many users, scripting takes the everyday drudgery out of repetitive tasks, letting them focus on more interesting things: building models, designing and running experiments, and finding useful ways to communicate results. For others, scripting extends JMP’s feature set, taking JMP beyond its native capabilities. New features and architectural changes in JSL include:
· faster list processing
· auto-saving
· better namespace control
· support for classes
