Experimenters' Club (prior DOE Club)

Dear Members,  We had the pleasure to welcome Chris Gotwalt (Chief Data Scientist, JMP) during the last Nordics Users Group Meeting.  His talk was recorded and you can watch it here along with the presentation slides: https://community.jmp.com/t5/Nordic-JMP-Users-Group/Presentation-Fast-Flexible-Innovation-with-Bayesian-Optimization/m-p/972729   Happy Friday!  ... View more

Video 1: Question on Bayesian Optimization use cases Victor: Maybe we can share the acquisition functions in the platform and start explaining them? Yuliana: I am observing that agentic ai propose kind of surrogate DOE + gaussian process as active learning workstream of experimentation. I still have impression that BayesOpt on JMP performs better in case we want to go faster, and I still can not understand why... probably there are more acquisitions functions. Answer: Depends also on the configuration of the Gaussian Process, and the acquisition functions worked by the JMP team are effective.   Video 2: Possible use case to use Bayesian Optimization Victor: Don't hesitate to mix approaches: start with small DOE's and continue with BO for example. Might not be appropriate with blocking or blocking with iteration. Links: Overview page: Bayesian optimization: Learn what Bayesian optimization is and how it helps scientists and engineers balance complex goals and find optimal solutions with fewer experiments. Bayesian Optimization Batch Customizer : here are details about the batch customizer options and the different acquisition functions as Victor is explaining Marketplace extension: Bayesian Optimization Simulator for JMP® Pro  Video 3: Constrained DoE - disallowed combinations, restrictions on responses, covariates Victor: If you don't have hard to change factors and you have multiple runs per batch, maybe using the "iteration" column as a block when running a regression model Can you change the factors definition? Instead of having X1 and X2 and a constraint on the ratio, use the ratio and total quantity in your factors? Yuliana: Is the ratio a linear constraint (system of 2 linear equations)? Sometimes I use responses that you can get quickly as factors (in place of real factors) to modelize more complex response  Links: Discovery Summit talk: https://community.jmp.com/t5/Abstracts/Synthetic-Chromatograms-A-New-Approach-to-Chromatographic/ev… Case Study: Functional Data Analysis for HPLC Optimization Talk from Frank at the Benelux user Group: https://community.jmp.com/t5/Benelux-JMP-Users-Group/Formulation-Mixture-Selection-to-Model-Relationships-Between-NIR/m-p/955063   ... View more

Hi all, I missed this week's meeting when I wanted to mention this 2-3 pm US Eastern Time Mastering JMP.  It is a first by @wdcorley1 From Experimentation to Optimization.  See how to combine experiment design with predictive modeling to support process understanding and optimization. Learn how to design an experiment, explore factor-response relationships, assess tradeoffs, and identify settings that achieve desired performance objectives. ... View more

The Experimenters' Club is shaped by its community. We'd appreciate your feedback on future topics and session formats. Your ideas will help us make upcoming meetings even more relevant and useful.   Survey link: https://forms.cloud.microsoft/r/S2kyz1bTwC   Thank you in advance!  ... View more

Video 1: Challenge -  Beyond One Best Model: What a DSD Can Really Tell You The Q2 session of the Experimenter’s Club featured a lively discussion on George Box’s famous reminder that "all models are wrong, but some are useful." It began with a user questioning whether unexpectedly strong interaction and quadratic effects in a 6-factor Definitive Screening Design (DSD) should be trusted. What followed was a thoughtful exchange showing that the real issue is not whether one model is “correct,” but how to interpret a design space in which many plausible models coexist. Community Question: https://community.jmp.com/t5/Discussions/quot-Surprising-quot-results-in-an-DSD-Design/td-p/941518 LinkedIn post from Victor: https://www.linkedin.com/posts/victorguiller_experimentersclub-designofexperiments-share-7453440558682894336-j7Ki?utm_source=share&utm_medium=member_desktop&rcm=ACoAAAg2XG8BduAUlMeluDXcuMFV-P_Hxv1KA0k Video 2: Discussion The conversation highlighted that in a DSD with a limited number of runs, it is impossible to estimate every main, interaction, and quadratic effect simultaneously without structural tradeoffs. As a result, partial aliasing, multicollinearity, and model multiplicity are natural features of the problem rather than signs of failure. The discussion emphasized that different estimation methods and selection criteria, such as R², adjusted R², AICc, BIC, and RMSE, can lead to different but still defensible models. A key takeaway is that experimenters should not focus only on finding a single best-fitting model. Instead, they should compare multiple strong candidates, look for effects that appear consistently across good models, apply effect hierarchy and domain knowledge, and consider augmentation when confirmation is needed. In that sense, the design process becomes less about reacting to surprising results and more about disciplined model judgment in real-world experimental work. ... View more