A more complicated model is seen using Borehole Latin Hypercube.jmp, found in the Design Experiment folder.
To launch the analysis, fill out the Gaussian Process dialog as shown in Borehole Latin Hypercube Launch Dialog.
When you click OK, the following Actual by Predicted plot appears.
Since the points are close to the 45 degree diagonal line, we can be confident that the Gaussian process prediction model is a good approximation to the true function that generated the data.
The Model Report shows us that this is mainly due to one factor, log10 Rw. The main effect explains 87.5% of the variation, with 90.5% explained when all second-order interactions are included.
Factors with a theta value of 0 do not impact the prediction formula at all. It is as if they have been dropped from the model.
The Marginal Model plots confirm that log10 Rw is a highly involved participant in Y’s variation.