Widely understood computerization and the continually growing value of information in the aspiration for creating a modern "informative economy", as well as advanced computer techniques, enable to systemize and unify the economic and social processes often presented in the form of developed statistical and mathematical models. Building statistical models involves many stages, and one of the most important among them is the selection of model variables. This paper presents a preliminary method for variable selection with non-classical assumptions while developing a regression model, connected with non-linearity of the phenomena taking place on the real estate market.

71-81