Of course, all Nonlinear Estimation examples were textbook examples, which usually have the tendency to work well. In actual research applications, you will surely come across data sets that seem to "actively reject" the model chosen for them. Refer to the Introductory Overviews for a review of some strategies that you can follow if a model cannot be fit to the data.

In general, it is always a good idea to begin with the simplest model possible. Also, the nonlinear regression techniques presented here can be considered hypothesis testing procedures, thus, they should generally not be used for exploratory data analyses. Rather, the researcher should bring to the data a prior understanding of the underlying mechanisms that relate the variables of interest.

STATISTICA includes various modules for fitting nonlinear models to data; some of these methods are specifically optimized for particular classes of (nonlinear) models, and hence, are much more efficient during estimation, and comprehensive with regard to the reported (specialized) results. For example, see Generalized Linear Models (GLZ) (which includes methods for Probit and binomial and multinomial Logit models), and Generalized Additive Models (GAM).