Can I add error values to a response variable in Custom DOE?

[#1]

Hi there,

I'm working with Custom DOE, and I've collected data for my response variable that I want to input to generate a model. I've also collected data on the error associated with the measurement of that response variable (for each data point) and I was wondering if there is a way to take this error into account. I haven't been able to find any details on this so I figured I'd ask here -- I'm using JMP Pro 17.

Thanks in advance!

[#2]

First, welcome to the community.  Interesting question.  Here are my thoughts.

My first thought is how much measurement error is there?  It may depend on how much measurement error there is.

Can you use averages to reduce the error?  I suppose you could use different values of the extremes of the measurement error distribution as different Y values (e.g., lowest, largest and midpoint)  and model each to see how that impacts the significance of model effects.

Usually you design an experiment with a model in mind and then as a result of the analysis, you reduce the model by removing insignificant terms and then iterating.

Perhaps others will have some ideas for you.

[#3]

Can I ask why you want to 'take this error into account."? It seems to me the measurement error is baked into the responses that have been observed already. It's gonna end up in your mean square error in the model you ultimately specify. Or are you worried that in certain parts of the experimental space measurement error is problematically large? If that's the case you can model the log of the variance via the Log Linear modeling capabilities. You'd end up with two models: 1. For the mean of the response. 2. Log variance of the response which is really just the measurement system error. Here's a link in the JMP documentation for the log linear variance modeling capabilities: Loglinear Variance Modeling in JMP