User`s guide

8 Model Analysis
When a low-o rde r mo de l fits the validation data poorly, try estimating a
higher-order model to see if the fit improves. For example, if a M odel Output
plot shows that a fourth-order model gives po or results, try estimating an
eighth-order model. When a higher-order model improves the fit, you can
conclude that higher-order models might be required and linear models might
be sufficient.
You should use an independent data set to validate your models. If you
use the same data set to both estimate and validate a m odel, the fi talways
improves as you increase model order, and you risk overfitting. Howe ver, if
you use an independent data set to validate your models, the fiteventually
deteriorates if your model orders are too high.
High-order models are more expensive to compute and result in greater
parameter uncertainty.
Nonlinearity Estimator Produces a Poor Fit
InthecaseofnonlinearARXandHammerstein-W iene r models, the M o del
Output plot does not s how a good fit w hen the nonlinearity estimator has
incorrect complexity.
You specify the complexity of p iece -wise-linear, wavelet, sigmoid, a nd
custom networks using the number of units (
NumberOfUnits nonlinear
estimator property). A high number of units indicates a complex nonlinearity
estimator. In the case of neural networks, you specify the complexity using
the parameters of the network object. For m ore info rmation, see the N eural
Network Toolbox doc umentation.
To select the a ppropriate complexity of the non linearity estimator, start
with a low complexity and validate the model output. Next, increate the
complexity and validate the m odel output again . The model fitdegradeswhen
the non linearity estimator be comes too complex.
Note To see the model fit degrade wh en the nonlinearity estimator becomes
toocomplex,youmustuseanindependentdatasettovalidatethedatathatis
different from the estimation data set.
8-68