First Time User - Custom DoE

[#1]

Hello,

I have done design of experiments before, but pretty standard ones (screening design, full factorial).

I have never used JMP before and never attempted at trying my hand at a custom design. I was hoping I could get some input as to the viability of the custom DoE I have created.

I have 4 factors (time, temperature, amount of catalyst, solvent) plus one co-variate factor (water content in my solution).

I want to see how these factors affect the molecular weight of the polymer I am synthesizing.

Any input is appreciated!

 

[#2]

I am not sure if I can answer your question regarding viability.  Also can't answer if this would be your most effective or efficient means to accomplish what you want.  Just not enough information about the situation.  There are lots of options. 

Custom design is based on the user having some idea of what effects need to be estimated and which do not (and perhaps are there any specific criteria you want to investigate).  I don't know what you predict, but it appears you are concerned with some non-linear effect as it looks like you have some pseudo center points in the design (they aren't really centered). You can evaluate the design using the green arrows under the custom design.  My guess is you have some randomized replicates which may be useful for allowing you to estimate a mixed model.

[#3]

Hello statman,

Thank you for your response! I do have an idea of what effects I need to estimate. And as for the center points and replicates --- I added 2 of each while making the design as I've been often told that this makes the overall design better.

Thank you!

[#4]

Hi Tehol-

Welcome to JMP!

The Custom DOE platform is a great place to start for a DOE and is appropriate for the experiment you described. Here are a few comments/observations:

1) It appears the Polarity factor is continuous, but the values do not seem quite right. Typically, the centerpoint value of a continuous factor should be halfway between the min and max values. In this case, you have a minimum of 2 and maximum of 38, so the midpoint is 20 (but "7" appears in the table). If the polarity values are due to the choice of a different solvent, you might consider making "Solvent" a categorical factor. You can do this in the Custom DOE platform, but when you add the factor, select Categorical and designate how many levels in the factor (2 for two solvents, 3 for three solvents, etc.).

2) Based on the fairly wide factor settings, this appears to be a screening DOE where you might be trying to uncover main effects. You have enough experiments here to start looking at interactions and polynomial terms, however, which might be more appropriate for an optimization DOE. The only downside to this is that you may be running more experiments than you need to initially run, especially if you are exploring factor settings that you typically do not run or do not have much historical support.

3) On a practical note, if there is a way to avoid the covariate factor this would simplify the experiment (by drying the solvents, using single batches where the water content does not vary over time, etc.). If the covariate factors are unavoidable, I suggest you read this excellent blog post on covariate factors in the JMP DOE platform: 

https://community.jmp.com/t5/JMP-Blog/What-is-a-covariate-in-design-of-experiments/ba-p/361517

I hope this helps and good luck with the DOE. Would love to see the results after you run the experiments!

-Scott

[#5]

Hello Scott,

1) Yes, that is a good point! I am using three different solvents. My main interest is to see how the polarity of my solvents will affect my reaction. I was under the impression that making it a continuous factor would be better. That is why I opted for the 'discrete numeric' option when setting it as a factor. I will revert it back to a categorical factor!

2) I was initially going to use a definitive screening design but I couldn't incorporate all the factors and levels I desired which is why I opted for a custom design.

3) I cannot avoid the water content, sadly, due to the nature of experimental setup I have. It is also known to influence the reaction so I thought it best to add it as a co-variate. I will definitely check out that blog post!

Thank you!

Tehol