Fundamentals of Statistics contains material of various lectures and courses of H. Lohninger on statistics, data analysis and here for more.

Exercise - Modeling the NOx Exhaust of an Experimental Motor

One of the goals in motor design is to keep exhaust as low as possible. The data set ETHANOL contains data on the NOx concentration which is to be modeled by the equivalence ratio (a measure of the fuel/air mixture), and the compression ratio. 88 observations have been taken from an experimental one-cylinder engine .

Go to the  DataLab  and try to model the relationship between the compression ratio, equivalence ratio and NOx concentration.

Note: this data set is especially difficult to be modeled by linear methods, so you may want to use a neural network.

  • Try to establish a model based on an RBF neural network, using the following parameters: S=0.1, R=0.0, Expol=2
  • Vary the number of hidden neurons between 2 and 40 and observe the results.
  • Perform a cross-validation (using a 2-sample step) of the results and plot both the PRESS value and the goodness of fit against the number of hidden neurons.