Summary of Statistical and Mathematical Techniques Used in State Energy Modeling
The basic statistical tool used in the first phase of most energy forecasting models is regression analysis. This class of techniques permits the modeler to extract from the historical record what might be thought to be measurable causal forces, and to assess their individual and aggregate impact upon the energy characteristic under consideration. The basic solution sought through regression techniques is the estimation of the weight or proportionality constant that links the measurement of two phenomena together. This chapter describes several tests to determine the role chance plays in the calculated value of the estimator. These include the T-test, hypothesis testing, Durbin-Watson test, and the F-test. In the historical estimation aspect of the energy forecasting models, concern focuses on the response of the dependent variable, usually an energy consumption term, to a change in an independent variable. Commonly, this response is measured as an elasticity. The chapter outlines several ways of calculating the elasticity.