nav[aria-label="Primary Navigation"] { padding: 0; & ul { list-style: none; width: 100%; display: flex; flex-direction: row; justify-content: start; align-items: start; gap: 30px; padding: 0; & li { margin: 0; } & ul li { list-style: none; } } }

JMP and Predictum- Design of Experiments DOE using JMP

najem

Please assist me with using JMP for design of experiments and predicting experiment results to minimize the number of times to conduct the experiment.

Im basicially looking to design an experiment to inject fluid concnetration at a certain rate into a rock and measure the output concentration.

Accepted answers

gail_massari

Hi,

As people have indicated, we try to make designing an experiment using JMP software user-friendly building upon the users understanding about designed experiments. I can't tell your level of experience with designed experiments from your post. Perhaps these Mastering JMP sessions and resources for people who use JMP already might be useful to give you a basis for using the software.

Basic DOE - broken into several videos

Mixture Designs - broken into several videos

Definitive Screening Designs - broken into several videos

Also, the April 6 live 1-hour webcast on mixture designs is open for registration.

All comments

cwillden

Hi najem,

I'm not sure this is the right venue for this kind of request. There is a lot of considerations that go into designing an experiment, and it would really require a consultation with a statistician or an experienced DOE practitioner to determine an appropriate solution for your specific situation. That's usually something you would pay someone to help you with.

I'll give you some free advice though. If you just have a single factor, injection rate, and your objective is to have a predictive model out the other end, you can probably just choose 3 levels equally spaced. This will allow you to fit a quadratic model, so you can model any curvature in that design space. You'll want to replicate those factor levels at least once each, especially the center point. How many times you should replicate those points depends on how noisy your response is. I would say you should have at least 8 runs so that you have 5 degrees of freedom for the error term.

Peter_Bartell

To add to the advice contained in @cwillden's reply, I also suggest utilizing JMP's Custom Design platform, Design Evaluation, and Compare Designs subplatforms. It seems in your original post, the number of runs is a mission critical issue. The Design Diagnostics -> Design Evaluation and Compare Designs platforms in JMP version 13 are tailor made for evaluating the effects of varying sample sizes on design performance characteristics like power and others.

gail_massari

Hi,

Basic DOE - broken into several videos

Mixture Designs - broken into several videos

Definitive Screening Designs - broken into several videos

Also, the April 6 live 1-hour webcast on mixture designs is open for registration.