One of the key steps in regression type modelling is variable selection. In the simplest case, the involved quantities and the functional relationship between them are known up to some unknown parameters which are to be extracted from the data. In general however, the necessary variables must be selected from a set of potentially useful variables and, if needed, be transformed. The selection is usually based on statistical criteria like Akaike’s information criteria (AIC). Major data analytical challenges are in situations, where the resulting model must be well-balanced with respect to data analytical and subject matter demands. A pure mechanical application of selection criteria rarely leads to a satisfactory solution. – How quantities might be selected, suitably transformed and how their relationship might be investigated in practise, will be outlined by examples from air quality management, from financial analysis and from health-economy.