Multivariate Regression: A Very Powerful Forecasting Method

Abstract

Regression Analysis is at the center of almost every Forecasting technique, yet few people are comfortable with the Regression methodology. We hope to greatly improve the level of comfort with this article. Here we briefly discuss the theory behind the methodology and then outline a step-by-step procedure, which will allow almost everyone to construct a Regression Forecasting function for both the linear and Multivariate case. The Linear Regression is shown to be a special case of the multivariate problem. Also discussed, in addition to model formation and estimation, is model testing (to establish statistical significance of factors) and the Procedure by which the final regression equation is obtained from the estimated equation. The Final Regression Equation is retained and used as the forecasting equation. A hand solution is derived for a relatively small sample problem, and this solution is compared to the MINITAB-derived solution to establish confidence in the statistical tool, which then can be used exclusively for larger problems.