Described in the context of two different time series models—(1) Traditional and (2) Box-Jenkins. In the traditional context, the MA model averages sales of certain periods to make a forecast for the next period. If, for example, we use a two-period moving average, then the average value of the last two periods becomes the forecast of the next period; if we use a three-period moving average, the average value of the last three periods becomes the forecast of the next period; and so on. How many periods of average to use depends on how far back the data seem to affect the sales of the forecast period. If you believe that the next period sales will be affected only by the last three periods, and then use the moving average of last three periods, and so on.

Previous Entry
Model Identification
Next Entry
Multicollinearity