Similarity between curves in Functional Data Explorer?!Hi all,
a customer is quite happy with FDE in general. They are now looking into options to compare curves and develop measures to decide whether two curves can be considered similar or not.
One application would be in scale up: Determine whether a production process performs similar to the lab scale.
I found some article (https://academic.oup.com/bioinformatics/article/27/22/3135/194486) but wondered if anybody has experience using JMP?
Best
Florian
Re: Similarity between curves in Functional Data Explorer?!Great idea Mark, The limitation I found with this is there is no way to save the integrated error from the target for each curve. I think saving that error would be the easiest way to do this.
Re: Similarity between curves in Functional Data Explorer?!What about using the target function feature in the FPC profiler? From the documentation for FDE:
FPC Profiler
Shows or hides a profiler of the FPC scores. The FPC Profiler includes a column for the input variable and a column for each FPC score. If a target function is specified, there are additional button options above the profiler graphs. You can optimize the target function and show or hide the target profilers. If you select Show Target Profilers, two additional profilers are added to the report. One measures the difference from the target function, and the other measures the integrated error from the target function. For more information about FPC Profiler red triangle menu options, see Profiler in Profilers.
Re: Similarity between curves in Functional Data Explorer?!Hey Florian, We have run into this a few times. The process we came up with, maybe fit this case.
Step 1. Make an FDE model with a set of "calibration/lab" runs
Step 2. Determine Specs on the FPCs that are considered "similar". This can be done with the profiler to see what the FPC changes do to the overall shape. Save out Spec Limits on FPCs. (*this might not work for the regulated industries)
Step 3. Bring in additional/production data to the "calibration/lab runs" and label that data as validation (This will prevent changes to the underlying model)
Step 4. Run process capability on those production runs and flag runs not in spec as not "similar"
Not sure if that accomplishes what you are looking for or not.
Re: Similarity between curves in Functional Data Explorer?!Thank you @malcolm_moore1 and @Steph_Georges , you are both right and these could be solutions. But as Malcolm mentioned already, I wish to stay within JMP Pro. I think this could generally be something people might wish for, especially in the pharmaceutical/regulated space, where they need to show similarity.
So again, thanks. I am still hoping for ideas with/within FDE.
To maybe spark more ideas, one thought is also to load a target and then have a "difference from target measure".
Re: Similarity between curves in Functional Data Explorer?!I do not know if it helps, but in Fit Curve > Dissolution Curve > Model Comparison, F2 Analysis and Multivariate distance are metric used to determine if 2 curves are similar or not.
More details in the Quick Start of the JMP Training Starter
Re: Similarity between curves in Functional Data Explorer?!Hi Florian,
I'm wondering what would happen if we treated scale as a supplementary factor in FDE with two levels (R&D and Scaleup). Then using Functional DOE with GenReg option. This would at least keep within the FDE platform if this option works. Otherwise you might need to explore Fit Curve which would give the option to compare curves for R&D vs Scale-up providing you can find a curve to fit appropriate to the data. Although this takes you away from JMP Pro.
Malcolm
