Attribute features application in specification of regression model of apartments cost

In the study of statistical data with a pronounced endogenous variable, it is necessary to identify factors (explanatory variables) that have a strong impact on the result. In this, factors can be both quantitative and attributive. To assess influence of numerical features, regression analysis methods can be used. Influence of attributive features is not taken into account. However, often these are they who make the decisive contribution to variation of the result. It is necessary to develop methods for analyzing influence of attributive features and accounting for these features in regression models. On the example of sets of apartments proposed for sale in the city of Krasnoyarsk, a new method is used to assess influence of attributive features on the quantitative using ranking them in accordance with their influence on the endogenous variable. Method of fictitious variables is used to analyze the attribute features. Each attribute with m values is assigned (m-1) dummy variables and a regression model is constructed. Influence of exogenous variables can be expressed using standardized regression coefficients. In this case, influence of attributes can be estimated by cumulative correlation coefficient calculated on the basis of a regression model with fictitious variables. For further research, set is proposed to rank, assigning each element a "rank" - value of a standardized coefficient which reflects closeness of the relationship with the endogenous variable. Thus, all features have a numerical value. A standardized regression model is constructed. Proposed approach can be used in the analysis of statistical aggregates, units of which are characterized by quantitative and attributive features.