The topic is strongly motivated by earlier work for the Ford Motor Company, which has already been successful and led to the 'Model-based Calibration Toolbox' for MATLAB produced by The MathWorks. There are further possibilities for research in the area, not necessarily restricted to automotive applications. The original automotive work focused on properly modelling the torque output of an engine in terms of variables such as spark advance at the point of combustion, speed of the engine, the air-fuel ratio, and the exhaust gas recycling ratio. The collection of the data meant that torque output was measured for a range of spark advances at specified values of the other variables. This induces a special data and error structure which requires innovative modelling. The earlier project mainly concerned torque as the output variable of interest, but there are others, such as the emissions NOx, CO and HC, which will require fuller and joint modelling. There are novel statistical concerns of more general interest here which need to be addressed, such as the assessment of selective aspects of fit, not overall goodness of fit; there may be areas of a fitted model which are not of relevance to the application of current interest. Never-the-less future projects are likely to be very much be embedded in engineering concerns and have strong links to the automotive industry.