Functional Data Explorer: applicable to study impact of a non-constant variable on a function?

[#1]

Hello,

We would like to understand what may have caused a change in aspect of an online process data curve over several runs.

The variable that we have identified as having changed over the past runs was the stirring speed. This variable however is not constant over time, but is varying over time. It is part of a regulation cascade to maintain another variable constant (the dissolved oxygen). This means it is not controlled over time but dependent on how much oxygen is needed. All controlled parameters seemed to be the same in all runs.

Would we be able to use Functional Data Explorer or another tool in JMP to investigate on the changes we've been seeing?

Many thanks in advance for your support!

Best regards,

Claire Baril

[#2 · Accepted solution]

Hi Claire,

Yes, the Functional Data Explorer should be able to do this, assuming you have a record of the uncontrolled stirring speed. The attached file has a sample script attached that shows an example of using parameters to model the curve (yellow in screenshot below)

[#3]

I will expand the suggestion by @Jed_Campbell. You have a set of output functions for the process runs and the corresponding stirring speed function for each run. You make a set of functional principal components (fPC) for the online process data curves, save them, and use them as your response in the Y role. You also make a set of functional principal components for the stirring speed functions, save them, and use them as your predictor in the X role. You need to be able to match each Y fPC with its corresponding X fPC for each process run. This arrangement is a matter of using commands in the Table menu. You can use most of the fitting platforms to propose and test models of a relationship between the stirring speed and the process curves.