Recording Experimenters' Club Q1 2026

maria_astals — Wed, 28 Jan 2026 12:55:25 GMT

Video 1: JMP's DOE platform: A new addition to the artist's toolkit

Read the blog from @BillFish828 here: https://community.jmp.com/t5/JMP-Blog/JMP-s-DOE-platform-A-new-addition-to-the-artist-s-toolkit/ba-p/919860

Question: What is the difference between Taguchi design and Orthogonal array.

MOLS is a balanced fractional factorial design matrix that ensures independence between factors, whereas a Taguchi Design utilizes these matrices to optimize the S-N ratio mean response enhancing robustness

Video 2: Why randomize when n=1

(view in My Videos)

Comment: My point was less about the trend, it was more about a not identifiable effect, so something you really did not know before the experiment. And my recommendation is to start with as many influences you can think of, even some far fetched influences, to select the factors from, and then observe any not included factor if possible.

Video 3: In-depth analysis of DOE's such as Response Surfaces (RS)

(view in My Videos)

Re: Recording Experimenters' Club Q1 2026

Victor_G — Sun, 19 Apr 2026 06:59:45 GMT

Thanks for the recordings @maria_astals !

On the topic of randomization, I don't totally agree with the purpose of randomization as avoiding time trend in the response. Randomization is there to avoid any bias in the experiment resulting from the influence of some extraneous unknown and uncontrollable factors that may affect the experiment. It is used to avoid confounding between treatment effects and other unknown effects (spurious correlations).
Depending on how the experiments are randomized, the design robustness against time may be more or less effective. Some better options (depending on how known and controllable the nuisance factors are) are :

Recording the actual run order and use it as a covariate in the model to evaluate its statistical significance and practical importance (again a "post-mortem" analysis to evaluate if order of the run may have an impact) for a known and uncontrollable nuisance factor,
Using blocking as @statman said, to balance factors levels as homogeneously as possible between blocks, in order to reduce or even cancel out the known and controllable nuisance effect(s). Blocking allocates similar runs in different blocks/groups of runs : different operators, equipment/machines, possible difference between days, between weeks, etc...
Design upfront the DoE by taking into consideration this possible time dependance using time as covariate, to balance the factors levels across time. @bradleyjones wrote a really good article on how to create a design robust to time-trend: How to create an experiment design that is robust to a linear time trend
You can see an application here:
The situation of using time as covariate to create a robust time-trend design is explained in the book "Optimal Designs of Experiments : A Case Study Approach" from Peter GOOS and Bradley JONES (Chapter 9: Experimental Design in the presence of covariates).

I wouldn't worry too much about one data point showing outlier/anomaly values, DoE are by design less sensitive to few outlier values than OFAT or other similar methods.

Looking forward to joining you next time !

topic Recording Experimenters' Club Q1 2026 in Design of Experiments Club Discussions

Recording Experimenters' Club Q1 2026

Re: Recording Experimenters' Club Q1 2026