Abstract
Modelling non-linear relationships: theoretical aspects. In many cases, the relationship between dependent and independent variables can be considered to be linear, assuming that the equation of the line from which the estimators are obtained is valid. However, there are situations in which the linearity cannot be maintained in the relationships between the variables. Thus, other non-linear functions must be used to represent them. The non-linear strategy of modelling allows us to perform, in an interactive way, the identification, estimation and validation procedures as well as the application of the models. From this point of view, it is possible to cover a large number of possibilities in function of the different characteristics of the variables involved. In this way, different models, such as those involving fixed independent variables, random independent variables, qualitative variables or autocorrelated errors can be used. This paper presents some theoretical aspects which should be taken into account during the process of fitting non-linear models, whose aim is to obtain valid relationships between variables.
