topic Randomization with Blocks in Discussions

Randomization with Blocks

NominalGemsbok3 — Fri, 17 Oct 2025 14:24:13 GMT

Hello

I have a Definitive Screening Design and, as JMP suggests, I selected blocks with center points to estimate quadratic effects. This results in two blocks (Block 1 and Block 2). Now I want to randomize. JMP suggests randomizing only within the blocks. I believe this is correct. However, the blocks are then not interleaved during execution (e.g., 1 1, then 2 2, then again 1 1, then 2 2, etc.), but instead, all runs from Block 1 are performed first, followed by all runs from Block 2. Is that still acceptable?

Regards Klaus

Re: Randomization with Blocks

statman — Sat, 18 Oct 2025 15:55:56 GMT

Blocking is a strategy to handle noise in experimentation. It accomplishes 2 things:

1. Increases the inference space, hopefully allowing conclusions to be extrapolated into the future

2. It increases the precision of the experiment.

This is because the noise confounded with the block is constant within the block (reducing the effect of noise), The noise constant within the block is then purposely changed between the blocks and is therefore assignable to the block effect.

Randomization is a strategy to handle noise that has not been identified by the experimenter.

"Block what you can, randomize what you cannot." G.E.P. Box

Re: Randomization with Blocks

Victor_G — Sun, 19 Oct 2025 06:19:35 GMT

Hi @NominalGemsbock3,

Blocking is a technique for dealing with nuisance factors.

A nuisance factor is a factor that has some effect on the response, but is of no interest to the experimenter; however, the variability it transmits to the response needs to be minimized or explained.

Blocking is the arranging of experimental units that are similar to one another in groups (blocks). The intent of blocking is to prevent large differences in the experimental units from masking differences between treatment effects, while at the same time allowing the treatments to be examined under different experimental conditions.
Practical situations involving blocks could be the use of different batches for raw materials, different operators, devices, or a practical constraint for running the experiments (for example, being able to run only 4 experiments per day). In all these situations, the use of blocking enables to have similar blocks of experiments, that help take into account this "constraint" without reducing the possibility to detect the effects of interest.

So blocking will force a specific order for running the experiments, to reduce the nuisance factor(s) behind it. You should not randomize the order of runs between blocks, the randomization is only set within blocks.

To learn more about blocking:
https://www.stat.purdue.edu/~zhanghao/STAT514/Lecture_Notes/LectureNotes13-Complete-Block-Design-.html
https://online.stat.psu.edu/stat503/lesson/4

Hope this complementary answer will help you,

Re: Randomization with Blocks

NominalGemsbok3 — Sun, 19 Oct 2025 07:53:10 GMT

Thank you for the answers. In general, I do understand the purpose of blocks, but I don’t understand how introducing blocks into a DSD is supposed to help detect quadratic effects more effectively. How does that work?

Re: Randomization with Blocks

Victor_G — Sun, 19 Oct 2025 12:58:03 GMT

Adding blocks with centre runs in a DSD assigns conference matrix foldover pairs to the same block. This way, the block effects are orthogonal to main effects, and the assumed model can estimate all main effects and quadratic effects.

Moreover, a centre point is added for each block, which enable to have a "control condition" that can be used to compare easily any deviation from block to block (beside the rest of the runs), and enable to increase the power of quadratic effects independently of the block effects (since they are distributed across all blocks). See

More infos here: https://www.jmp.com/support/help/en/18.2/#page/jmp/design-options.shtml#ww458084

Hope this answer will help you,