Does the model from the Test Plan at MBC Toolbox have an induction on the submodel created from the response?

Hi, while working with the MBC Toolbox I found something unexpected. I created two One-Stage Model's one with a quadratic model an one with an neuronal network model. Later then i build different submodells from the responses using the test data, the neuronal network submodels from the Testplan using the neuronal network model had significant better statistics than the other.

Does the model created in the process of making the one-stage model have an induction on the submodel's and if, how?

best regards, Tim

Answer 1

[Ian Noell]

Hi Tim,

I have just seen your question from last year. The best model type depends on the complexity of the response that you are trying to model. I would normally expect a non-parametric model such as a neural network or radial basis function model would fit engine data better than a quadratic. In R2015b, MBC also as Gaussian Process Models that have the flexibility of neural networks or radial basis functions but do not normally require user-configuration.

I do not understand what you mean by induction but would be happy to answer if you could clarify.

Best regards, Ian Noell