Basing Forecasts on Regression
IN THIS CHAPTER
Knowing whether to use the Regression tool
Getting familiar with the Regression tool
Taking the Regression approach
Regression is a standard technique in forecasting, whether sales revenues or sunspots. (And yes, meteorologists and astronomers have used regression for years in forecasting sunspots.)
This chapter introduces forecasting with regression. The idea is to get your hands on one variable (say, the price you charge for your product) that is strongly related to another variable (say, your unit sales), and then use what you know about the first variable to forecast what will happen to the second variable. Of course, “what you know about the first variable” ranges from the obvious (“Next month is April”) to the mysterious (“Is the Fed about to raise interest rates?”). Generally, you’d prefer to avoid a situation in which you find yourself forecasting your predictor variable.
That’s simple regression: One variable forecasts another. You can also make use of multiple regression, where you use more than one variable to forecast another. A typical example is to use both product price and the index of consumer confidence to forecast sales. Within limits, you can use as many predictors as you can lay your hands on — often (again, within limits) the more predictors you use, the more accurate your forecast. It’s important to keep in mind that the more predictor variables you have, the more records you need. A regression equation ...