ABSTRACT
The time series models of Chapter 5 and 6, and subsequent methodologies of regression (Chapters 7, 8, and 9) can be effectively used to make forecasts in most business and economic environments. However, these models do not always capture the pattern that may be exhibited by many time series. For example, methods such as the exponential smoothing are suitable for shortterm forecasting of time series. However, when faced with businesses and economic conditions that exhibit more complicated data patterns such as a combination of a trend, seasonal factor, cyclical, and random fluctuations, we need a more comprehensive method. Similarly, in using a regression model, the analyst depends on visual (scatter diagrams) and statistical analyses to determine the “best” model for forecasting purposes. The iterative process of determining which model is the “best” is expensive and time consuming. Furthermore, the utility of the regression models depends heavily on satisfying the assumptions of these models. Even when most of the assumptions are satisfied, we were not certain whether the model would provide a good forecast in the long run.
