Abstract

BOADA, Antonio José  and  DE VASCONCELOS, Diego. Multiple regression statistical model. Rev. Lasallista Investig. [online]. 2013, vol.10, n.1, pp.112-127. ISSN 1794-4449.

Introduction. By the use of this modeling procedure, creating a general statistical scheme was possible. Such model could be standardized for the company's whole products catalog, aiming to make simulations and predictions with a high confidence degree concerning the evolutions in time established by the same company. Objective. To show an innovative estimation system created to predict the demand of the products, which are segmented by their inventorial reference numbers, SKU (Stock Keeping Unit). Materials and methods. A statistical model was made to predict the products' demand in a time period determined by the company. A static multiple regression model was used, taking into account especial marketing variables for direct sales companies. Results. In the studies performed, R2 adjustment levels between 70% and 80% were achieved, with a minimal multicollinearity between variables and a statistical behavior of statistically acceptable residuals under randomness and normality terms, and with wide variance stability. Conclusion. This multiple regression model is an option that can be the base for prediction in catalog selling multinational companies.

Keywords : prediction; products' demand in companies; causal variables in products' demand.

        · abstract in Spanish | Portuguese     · text in Spanish     · Spanish ( pdf )