Currently in FDE Wavelet Modeling you are given the option of saving out all of the coefficient data to a new table and then using that data to build a predictive model from the full set or a subset of the coefficient data.
During the Wavelet DOE analysis there is an Energy table/graphic provided that gives the most influential coefficient data, and in the case of spectral data the energy values are for coefficients of individual wavelengths. I would propose that being able to select and use the higher energy coefficients (wavelengths), instead of using the full set of coefficient data would be faster and easier to interpret.
There was a potential issue uncovered in using a model built from the full coefficient table in that many of the higher energy terms were zeroed out (4 out of 6) and not used in the model. As shown in the image below the more complex model is only slightly better than the Energy Set Only model.
This is an internal request based on what is coming in the JMP Pro 17 FDE platform.
There is no competitive software that I know that does this type of analysis.
In my personal opinion only, this will enhance the wavelet analysis platform and it will help drive adoption of JMP Pro for the analysis of chemometric data.
