JMP’s latest release brings a fresh wave of updates to the Design of Experiments (DOE) platform, with new tools and smarter workflows that make it easier to design, simulate, and explore experiments. Whether you're screening factors, simulating responses, planning sample sizes, or trying to pinpoint the root cause of a failure, there’s something new to try out.
Smarter fault localization with BayesFLo
Among our DOE experts at JMP, one of our favorite topics is testing complex systems (including statistical software). But when you use a covering array and uncover a failure, you’re often left with a long list of suspicious combinations and no clear starting point.
Bayesian Fault Localization (BayesFLo) in JMP 19 gives us a smarter way to pinpoint root causes. Based on research Irene Ji began during her doctoral program and implemented in JMP with help from Joseph Morgan and Jacob Rhyne, BayesFLo adds a probabilistic twist to the Covering Array Analysis platform.
What’s new? Instead of just counting how often a combination shows up in failures, BayesFLo uses Bayesian inference to rank combinations by their likelihood of being the culprit. That means fewer ties and more clarity on where to start, taking advantage of your domain knowledge with a ranked list based on posterior probabilities.
Want to try it out? Head to DOE>Special Purpose>Covering Array>Analysis in JMP Pro 19 and look for the new Rank by Probability checkbox, which reveals BayesFLo options.
Power and sample size explorers get a boost
If you’ve used the built-in Sample Size Explorers in JMP before, you’ll feel right at home, but you’ll also notice some nice upgrades. Caleb King, with help from Yeng Saanchi and Jacob Rhyne, has built them right into JMP, which means they are now more robust and the reverse lookup functions are more flexible and capable of handling much larger sample sizes. You can:
- Click Set Targeted Response to solve for power or other variables more intuitively.
- Save your setup as JMPDOE files to share with colleagues or revisit later.
These updates make the Sample Size Explorers more flexible, more powerful, and easier to continue developing in future releases.
While you should give them all a try, there are some other new explorers you might also want to check out:
Parametric Tolerance Intervals Explorer
If you’ve ever wanted to explore the extra “buffer” a tolerance bound or interval contains based on sampling uncertainty, now you can do that directly in with the new Parametric Tolerance Intervals Explorer, available under DOE>Sample Size Explorers>Reliability.
Choose from such distributions as normal, lognormal, exponential, Weibull, and SEV, and see how sample size and distribution parameters influence the width or exceedance of your tolerance bounds. It’s especially helpful for non-normal distributions.
Nonparametric Tolerance Intervals Explorer
JMP 19 also introduces a new Nonparametric Tolerance Interval Confidence Explorer, designed to help you understand how sample size and order statistics affect the confidence level of a nonparametric tolerance bound or interval. Check out this explorer under DOE>Sample Size Explorers>Reliability.
Acceptance Single Sampling Plan Explorer
Planning a single acceptance sampling test? The new Acceptance Single Sampling Plan Explorer gives you a profiler-style interface to explore how sample size, number of failures, and defect rates interact, whether you're using lot sampling (hypergeometric) or binomial sampling. Set AQL and RQL values, solve for acceptance probability or risk levels, and view color-coded reference lines to guide your decisions. You’ll find it under DOE>Sample Size Explorers>Quality.
Smarter simulated responses
Simulating responses for a designed experiment is a great way to try out designs before collecting data. Thanks to some great teamwork from Mark Bailey, Joseph Morgan, and Jacob Rhyne, JMP 19 gives us a lot of flexibility for simulating responses. The new Enhanced Simulate Responses lets you simulate responses using random coefficients – a great way to simulate responses where you can investigate what happens to the responses (and modeling) when only a few of the coefficients are nonzero (and of different magnitudes). The random coefficients provide options to enforce effect sparsity, effect hierarchy, and model hereditythat underly our assumptions when designing experiments.
Previously, the Random Coefficients option in Simulate Responses was only available in Easy DOE and did not consider the structure of the model. Now, you can:
- Control how many effects are active (sparsity).
- Scale interactions and quadratics relative to main effects (hierarchy).
- Ensure that if a second-order effect is active, its related main effects are too (heredity).
You can also reset the random seed for reproducibility, initialize coefficients to a constant, and adjust active proportions for each effect type. This update makes it easier to simulate realistic responses based on sound statistical principles, whether you're teaching, testing, or just exploring your design. Available in most of our DOE platforms, the new options will show up when you have Simulate Responses selected when creating your data table.
Modern screening designs now in Design Explorer
Design Explorer has been a favorite for quickly comparing custom designs, and in JMP 19, it’s gotten a major upgrade. You can now explore definitive screening designs (DSDs) and other modern screening options directly within Design Explorer – no need to build multiple tables or jump between platforms.
This update makes it easier to create and compare DSDs with different numbers of center points or extra runs and explore near-orthogonal arrays (NOA) and orthogonal mixed-level designs (OML) that were previously tucked away in Classical>Factor Screening>Screening Design.
Just launch Custom Design, and if you have a main effects model with continuous and two-level categorical factors that will accommodate these designs, Design Explorer will provide these additional options. Even if you haven’t used it before, check out Design Explorer as a great way to explore trade-offs and properties across multiple designs all in one place.
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.