Hi, is it possible in JMP PRO to get Proven Acceptable Range from a OFAT study with factor varying over the time and multiple response variables to consider ? (responses values are fixed and not varying over the time) If yes, which functionality should I use ? Is this functionality available in version 17.2 of JMP Pro ? Thanks in advance for your help!
Hello Vincent,
In your situation, you have several X, and some of them (at least one) is dependant of the time. and you have several Y and you want to explore the relationship between them
Let's say something like Y1=f(X1(t), X2(t), X3).
In this situation, the Functional Data Explorer (JMP Pro) is really useful (and available in 17.2).
Fermentation Example:
For example, imagine that we have Y=Yield, and X1=Ethanol level (depending of the time) in a fermentation process.
You can set the platform as following:
You have to choose an appropriate model. In this example, B-Spline is good enough. Depending the data, wavelet could be a better solution.
and you have now established the relationship between the yield and the ethanol. We can see the impact of having high ethanol at begining or in middle of the process on my Yield.
Specific Situations: several X (including X non dependant of the time)
in this case, the relationship beteween the Xs and the Y is a little bit more complex and needs to be build outside the patform. You will have two steps:
Step 1: In this case, on the platform launcher, put every X dependant of the time in the Y,Output area and every X non dependant of the time in the Z,Supplementary area.
Once you've created the model for each time-dependent X on the platform, you need to extract the coefficients (Save summaries) into another data table. In this table, you'll have your Ys, your Xs (simple measurement) and for each time-dependent X, the calculated coefficients.
Step 2: It's then possible to create a model to create the relationship. I will personally use a LASSO model (manage correlation, and highlight only important factors)
The final model will give you the relationship between Y and the different X and Functional Principal Component. If you want to see the effect of the principal component on the curve, save the equation, and use the profiler from Graph menu with the option 'Expand intermediate formula'
In the following video, you can see that the second principal component of ethanol (in yellow) has effect on the yield. I need to minimize this coefficient to optimize my yield . But what does it mean on my ethanol curve? You can have a look on the shape of the ethanol yello curve when I change the amount of my ethanol FPC2,
Same for Molasse feed FPC2. I need to have a hive value of this component to improve my yield (in orange). And I can see the shape in good or bad situation on my orange curve.
Same in Red for my NH3.
Range:
On this model, we can simulate a variation on the different FPC, which have an impact on the yield but also on the shape of the curve and give you some acceptable range of variation.
Note that if you have a specific project, I will be happy to discuss with you about it.
Hello Florence,
Thank you very much for your answer!
