Hello,
I am using JMP Student Edition 19.1.4 to analyze functional data from a kinetic experiment.
My experimental design is based on a 3-factor Box-Behnken design. I know that the experimental factors have significant effects on the responses, but these effects can vary over time. Therefore, I am interested not only in the response at a specific time point, but also in how the kinetic profiles change depending on the experimental conditions.
I have 15 experimental runs, with measurements taken at 10 time points, from approximately 0.5 h to 24 h. I used Functional Data Explorer with a B-spline model and VCG to model the kinetics.
The model seems to fit the experimental data reasonably well, and JMP provides fitted curves for each individual run. I would now like to investigate whether it is possible to use these fitted functions to obtain estimated values at time points that were not experimentally measured, particularly between 7 h and 24 h.
As an additional analysis, I used the functional principal component scores (CPF) obtained from the Functional Data Explorer as responses in a Response Surface Methodology (RSM) analysis. This allowed me to investigate how my experimental factors affect the main features of the kinetics.
However, I am not sure what would be the most appropriate way to proceed from here. Ideally, I would like to obtain the fitted functional model for each individual run and use it to estimate the response at intermediate time points, rather than simply interpolating the raw observations.
I cannot find a way to export the continuous fitted function or obtain predictions at arbitrary time points from the Functional Data Explorer.
Is there a way to extract the fitted B-spline function (or the necessary coefficients) for each individual run in JMP Student Edition?
More importantly, would using the fitted B-spline functions in this way be a statistically appropriate approach for estimating values between the experimental time points? Or would there be a better way within JMP to obtain these predictions?
I would also appreciate any advice on whether my overall approach (using the functional model to describe the kinetics and then using the CPF scores in RSM) is appropriate for this type of experimental design.
Thank you for any guidance or suggestions.
