Innovative applications in designed experiments

[#1]

Last week’s Statistically Speaking was a first in that we had very few questions from our normally very inquisitive, very curious audience. Since we heard from a few attendees that they had difficulty submitting a question, we invite you to ask them here.  

For those of you who missed this episode, you can watch it on demand at your convenience, as well as see the slides from JMP presenters: Ryan Lekivetz, head of the Design of Experiments (DOE) and Reliability Development team, and Yeng Saanchi, Analytics Software Tester. They shared highlights of innovative applications in DOE and the first case study on a more strategic approach to software testing. The event also included former JMP intern, Bryce Scott, who explained how a designed experiment was used to optimize 3D printing in the aerospace industry.

It was a great opportunity for us to share how JMP engages with graduate research assistants, which is how Yeng came to be at JMP, as well as our amazing intern program, which is how we came to know Bryce. JMP really does celebrate our intern program, as reflected in social media posts like this one:

JMP interns aren’t coffee runners – they do real, meaningful work. Our award-winning internship program is focused on development, culture, and community. We’ll help you grow professionally, find (or further) your passion, and make memorable connections that last beyond the summer.  Interested in learning more about an internship at JMP?  Follow our Careers Page for Summer 2025 opportunities, which will be posted in September and October.  And check out this short video one of our interns made.

Since we didn’t get as many questions as we normally do, Ryan, Yeng, and Bryce share answers to questions they’ve had in the past when presenting on these topics. 

Does your group have any internships?

Ryan:  We usually post our internships at jmp.com/careers in October. Some postings will be for particular teams (like DOE and Reliability), while others may be for general interest in an internship.

Can covering arrays be used in other application areas besides software?

Yeng:  Certainly. Covering arrays are a generalization of orthogonal arrays in DOE that are used in fractional factorial designs. Fractional factorial designs are an economical alternative to full factorial designs without sacrificing a lot of information. Covering arrays are also used in hardware testing to detect failures due to interactions between components. Other areas in which covering arrays are used include genomics and digital communications.

In the field of genomics, covering arrays are used for compressing multiple sequence alignments. The alignment of genomic sequences is important to detect differences between a gene for one species and a gene for another species due to evolution. The number of genes to consider in such cases is so large that the cost of considering all possible alignments can be prohibitive. Covering arrays are used in such instances to ensure that each gene from one species is aligned at least once with each gene from another species. From a covering array perspective, every species is a factor and every gene for that species is a level of the factor.

In the area of digital communications, covering arrays can be used to resolve conflicts that may arise when multiple users are sharing a unique resource. In such situations, a conflict is said to arise when two or more users access the communication channel in the same time slot. Covering arrays are employed to solve this communication conflict by finding an available time slot for each user willing to send a message.

Did the important factors/settings have any meaning to you if tried to apply them to a different printer?

Bryce: Yes, they did, although the exact values and performance of this DOE is specific to my printer, the basic process of printing is the same for all printers. So, an important factor for my printer will still be important for another printer, maybe not to the same extent, but it will still give us some insight into what the important settings are. 

Have you tried another DOE since your internship?

Bryce: Unfortunately, not. I have tried to apply it to some of my lab experiments, but the work I am doing is often regulated and requires that we test each factor one by one. 

Do you have questions about this episode of Statistically Speaking? Or questions about innovating with designed experiments? We’d love to hear them. To see more examples of innovative applications of designed experiments, check out this episode of Technically Speaking, where another of our subject-matter experts, Tom Donnelly, shares how you can use DOE to reverse engineer formulations.