Explaining beer demand: A residual modeling regression approach using statistical process control

Abstract

We develop a medium-term model as well as a short-term model for understanding the factors affecting beer demand and for forecasting beer demand in Turkey. As part of this specific model development (as well as regression modeling in general) we propose a procedure based on statistical process control principles (SPC) and techniques to (1) detect nonrandom data points, (2) identify common missing, lurking variables that explain these anomalies, and (3) using indicator variables, integrate these lurking variables into the model. We validate our proposed procedure on several test examples as well as on the medium-term beer demand model. Both the medium and short-term models yield very satisfactory results and are currently being used by the company for which the study was conducted. In addition to the residual modeling regression approach developed using SPC, a major contribution to the success of the project (and the modeling in general) is the mutual collaboration between analyst and client in the modeling process.