Confidence Interval for Functional DOEA customer said they heard at Discovery that there is a trick to adding confidence intervals to a Functional DOE profiler. Was wondering if anyone knows how to do this.
@Peter_Hersh @clay_barker
Re: Confidence Interval for Functional DOEHi Andrea,
I believe they add a formula for the confidence interval to the Graph-Profiler and not within the FDE platform profiler.
However, I'm interested to know if there is another way or maybe included in a developers version.
Martin
Re: Confidence Interval for Functional DOE@martindemel Do you know which formula is used?
Re: Confidence Interval for Functional DOEHi Andrea,
I tried to look up the way they used, and had in mind there was a way to do this simpler. Unfortunately it seems they did a manual approach refitting each chromatogram in nonlinear platform. It was chromatogram data. Here are the notes I got:
- Smooth the raw data. this is really important , they used smoothing functionalities from the Graph Builder platform since the Functional Data Explorer has not a good feature at the moment for smoothing chromatographic data.
- Use the peak detection feature in the Functional Data Explorer to find the peak locations and peak heights of the two peaks of interest.
- Fit the parametric model in the non-linear platform by using the peak height and peak location information to get a faster and better fittings (still a lot of manual work to get a good fitting). Do this for each of the 40 chromatograms.
- Save the numeric values of all parameters and fit the DOE (i.e. Fit DSD)
- Combine the individual models into one model to generate the synthetic chromatogram
- “Hack” the Prediction and Confidence Interval Formula by adding the synthetic chromatogram model of only one peak, respectively. This allows for the overlapping of each peak component in the profiler (green and blue).
Not sure this will be of practical help, as it could be quite cumbersome "just" to get the intervals in the profiler. At least I would be very hesitant to recommend that approach.
Probably there has been made some enhancements already for v19/20? I'm with Clay and Ryan in Berlin this week, maybe they know more
