FDE in JMP Pro 14.1Hi,
I created a DoE and I have as response temperature. However, I measured the temperature through time so I would like to analyze the data with the Functional Data Explorer. The goal is to end up with the nice profiler which will return the trend of temperature and time similar to your example here: Using JMP® Pro to Model Functional Data looking like this:
However, I would like to have the FDOE profiler (or profiler close to FDOE ) in JMP Pro 14.1 which I have (and for technical reasons from my side I can’t change for now)
Thanks for your time.
DOEOtherRe: FDE in JMP Pro 14.1Have you read the documentation? There is an example with JMP Pro 14 on page 282 of the "Predictive and Specialized Modeling" user guide about how to proceed to model the functions with factors.
Re: FDE in JMP Pro 14.1Hi,
Indeed there is an example in 14.1. I think that it is a problem where you have data points(scalar) response and functions as predictors. There the data structure is in wide format.
On the other hand my data are more like the one in this example. The predictors are data points (from a DoE) and the response is a function. There the data are structured in long format.
I think that the process is clear till the Save Summaries stage (where I'm getting the FPCs) but then I don't know what to do (and I think there are the most differences in FDE between versions 14 and 15). I suppose to run a generalized model but I don't have the right choices.
Re: FDE in JMP Pro 14.1I conferred with JMP Development about your request. Here is their answer:
"In more recent versions of JMP ( I think JMP14.1 or .2) you can Save Functional Prediction Formulas from Generalized Regression. If not, you can save the prediction formula in terms of the FPCs from FDE and replace the FPC references in the formulas with the Generalized Regression prediction formulas"
Re: FDE in JMP Pro 14.1I've seen a video (below) showing in 14.2 how to save the Functional Prediction Formulas. But I can't see how to move forward with 14.1.
However, thanks for your time. The What, Where, Why and How of Functional Data Explorer ( 2019-EU-45MP-050 )
Re: FDE in JMP Pro 14.1@Elef, The FDE workflow improved dramatically between 14.1 and 14.2 enabling the ability to save functional prediction formulas. Also automatically generating profiler scripts in saved summary tables. In JMP 15 the workflow improves further with the addition of the functional DoE analysis built into the FDE platform. For JMP Pro 14.1 you can get the FPC scores and use them in modeling, but JMP did not have the ability to use a Z variable to pass through until 14.2. Meaning that the functional prediction formula had nothing to reference. You can see the impact of FPCs on shape, but not a DoE factor until you get to 14.2. Hopefully you will be able to update to a new version of the software soon.
Re: FDE in JMP Pro 14.1Hi,
I understand that there is no direct way to include the z variables however, I was wondering if there is a manual way to end up with a profiler similar to v15 (like the one in the picture).
Thanks for your time.Re: FDE in JMP Pro 14.1@Elef Unfournately not directly in 14.1. You are able to save out the FPCs and make a model using the FPCs as responses for your DoE factors. In 14.1 that model does not have the DoE factors so it is not able to generate the profiler shown at the top. However, this gives you a bit of a workaround as you will be able to see which factor settings lead to which FPC levels. Then you can manually change the FPC levels to the settings the model predicts from DoE factor settings (linking profilers won't work as FPCs are a factor in one profiler and responses in the other).
Re: FDE in JMP Pro 14.1Hi,
@phersh, I did what you suggested i.e. I modelled each FPC with the DoE factors (in total I have only two FPCs so, not that much work). However, the coefficients for the FPC1 model and the FPC2 model are totally different. All the factors show different direction. Now, what should I conclude? Should I stay with what the coefficients from the 1st model since the 1st FPCA explains like the 90% of the variation? Or there is a way to combine the coefficients from both models? I guess the 2nd option could give more robust results?
Thanks once again.
