The development of useful models requires appropriate methods but even more important are suitable data. For instance in a first trial, it may not be essential whether the regression method principal component regression or partial leastsquares or an artificial neural network is applied to make a calibration model, but it is essential that the used x-data have a strong relationship with the modeled y-data. Choice of x-data is often limited, because of the specific chemical-physical properties of the samples (objects), the availability of instruments, and also because we often only suppose that some x-data may be related to the desired y-data. Available x-data can be improved by appropriate transformation or extension-in general by an appropriate preprocessing. Some preprocessing methods are solely based on mathematical concepts, others are inspired by the chemical-physical background of the data and the problem. Selected preprocessing methods that are important in chemometrics are briefly described in this chapter. A book has been dedicated to wavelet transforms (Chau et al. 2004), other chemistry-specific data transformations are described in Brereton (2006), Smilde et al. (2004), and Vandeginste et al. (1998).