nav[aria-label="Primary Navigation"] { padding: 0; & ul { list-style: none; width: 100%; display: flex; flex-direction: row; justify-content: start; align-items: start; gap: 30px; padding: 0; & li { margin: 0; } & ul li { list-style: none; } } }

FDE Training

saradoudt

8.17.2020 | Presenters: Peter Hersh and Mary Loveless

Watch Recording | View Powerpoint | Take Quiz

What is FDE?

FDE is Functional Data explorer and makes analyzing shapes quick and easy. It turns large amounts of data into a more manageable form.

What is the difference between data and functional data?

Functional data are data used to describe curves, surfaces, or anything else that varies over a continuum… The physical continuum is often time, but may also be spatial location, wavelength, probability, etc. Sometimes the challenge of working with functional data is that often, scientists don’t want to analyze the functional data directly; rather, they want to work with the underlying functions which produce the observed data.

Value Statement

FDE allows Process Engineers and Statisticians to be able to analyze large amount of their streaming or sensor data. No one in the GUI environment has FDE. R and Python offer some of this functionality, however, JMP is the only out of the box solution that offers full FDE capabilities.

Ideal Audience

Almost every company can benefit from FDE. The customer should be statistically savvy. These are the types of folks you should look for when setting up an FDE JMP Discovered.

Common industries:

Semiconductor
Medical Device
Chemical
Oil and Gas
Manufacturing (CONTROL+F to find Edwards Vacuum)

Common roles:

Process Engineers
Industrial Statisticians/Engineers
R&D

Common situations/data:

Have Sensor or Streaming data
Repeated measurement data
Looking at Shapes or Patterns

Problems/Pain Points addressed by FDE

Better understanding of streaming data
Shape of data is important most customers are not analyzing this
Lots of measurements for a single batch with a lack of summarization

Terms/Phrases to Listen For

streaming
sensor
repeated measure
data collected over time
analyzes curves
analyzes shapes
spectral
dimensional reduction
mixed models
principal components
partial least squares
time series

Discovery Questions to Ask

If you had a tool that could analyze your streaming/sensor data for you, how much time would that save you?
Do you feel like you are only analyzing a portion of your data rather than all of it?
How are you using your streaming data every second/minute/hour, etc?

Action Items from this Training:

Using the industry/roles listed above, are there segments or accounts where you can/should position FDE?
Do you have leads in your territory that attended any of the below mastering JMP webcasts? These are great campaigns for an FDE followup.
Is there a potential in your territory to do an FDE JMP Discovered?

Related Resources:

Sales Stories where FDE was a factor (control+F to search for FDE)
Customer Story with FDE: FujiDiosynth
Customer Story with FDE: Symrise
Customer Story with FDE: Lynred
Book on Amazon on FDE
Blog Post: FDE in JMP 14
Blog: Optimizing Film Topology Using Functional Data Analysis
Blog: Chemical Reaction Monitoring Using Functional Data Analysis
Blog: A 5 Minute Introduction to Functional Data Analysis
Blog: Image and Wafer Analysis using Functional Data Analysis
Mastering JMP: Using JMP Pro to Model Functional Data
Mastering JMP: Tracking and Trending Manufacturing Metrics
Mastering JMP: Using Partial Least Squares Regression Just Won't Work
JMP Discovered (FDE)

Please refer to this page to find out what other training is available.

Comments

malcolm_moore1

Thanks for pulling this together Sara. It is particularly helpful to have links to the relevant resources. We have a lot of great information and stories to share and the challenge oftentimes is finding it.

Malcolm

tom_donnelly

Some additional trigger words suggesting Functional Data Explorer may be a solution are "Landmark Analysis." I have heard customers use this to refer to taking a shape and isolating key features such as peak height, distance between peaks, area under the curve, slopes of sections of the curve, etc. and then creating statistical summaries of these "landmarks." The summary statistics are then modeled to make predictions about the process. This discussion is relevant to the slide Pete showed a spectral distribution and identified the max, min and mean height of the peak, as well as heights of curve at specific wavelengths.

DonMcCormack

Repeated measures and mixed models. I'd be be careful if I heard these two terms because they can mean specific statistical approaches to data that are not related to FDA. If the focus is on statistical significance, then it's not a situation for FDE. In my experience, mixed models focus on a completely different aspects of data analysis than FDA. That said, the mixed model capabilities in Pro might be of interest in these situations. Repeated measures is a bit more ambiguous. Data taken repeatedly over time may be a good candidate for FDE if the analysis focuses on the shapes of the curves and not their statistical difference. Profile analysis is another term I've heard from people who are looking to do FDA.

Juliette_Plager

Great training, thanks! @saradoudt is this the page where the presentation will be posted? I'm specifically looking for Pete's "too much data" slide (PCA).

saradoudt

@Juliette_Plager - yep just added recording, ppt and quiz to the top of this page. Let me know if you need anything else!

Peter_Hersh

@DonMcCormack , Yes FDE is not a replacement for Mixed Models. Either way, they will need JMP Pro.

Tim_Pierce_JMP

@saradoudt , I saw a really nice one page infographic posted on LinkedIn, authored by Bernd Heinen @bernd_heinen . Title was "Linear is gone, Holistic Analysis and Optimization of Curves". I am guessing this is from a Discovery Summit poster/paper(I used the paper search add-in, but did not see it). I would love to be able to reference it in a link or as an asset. It does a great job of boiling the topic down into something than can be consumed quickly by users and just as importantly - users needing to sell it up the food chain internally.

If you find this in a shareable format, please let me know.

Cheers,

Tim

Briana_Codd

@saradoudt just wanted to let you know the two middle webinars in the Action Items from this Training section are not linked.

saradoudt

@Briana_Codd updated those links! thanks for letting me know.

martindemel

@DonMcCormack @Peter_Hersh
“have repeated measurement data” should probably be changed into something like this:

„repeatedly measure data over a continuum (time, frequency, distance, …)”

this may prevent raising wrong assumptions.

DonMcCormack

@martindemel in the context of FDA, I think the terms "continuous measurement data" or "continuously measured data" are better descriptors for the type of data that FDE is good for. Dropping the "repeated measurement" part makes it less ambiguous.

Peter_Hersh

@martindemel and @DonMcCormack great points both of you. The term repeated measurement has a very specific meaning to many people. I have heard the term tool trace data from a semiconductor customer and it was a great case from FDA/FDE.