Use Winters' multiplicative method to smooth the data in a time series, and to provide short to medium range forecasting. For example, you might want to predict the temperatures for the next 6 months using the data in a time series containing successive temperatures in a city over the past 48 months. Or you might want to predict the sales of a certain type of sport equipment over the next 4 months using data collected over the past 3 years (36 months).

Use Winters' multiplicative procedure when

·    You have data with or without trend

·    You have data with seasonal pattern

·    The magnitude of seasonal pattern is proportional to data, consistent with the model when the seasonality is multiplied by the trend

·    The series does not include any missing values

·    You want short to medium range forecasting

The procedure can be used when both trend and seasonality are present and are multiplicative.

You wish to predict water usage charges in a small school district. Bimonthly data were collected over the past 21 billing periods and these data will be used to make predictions over the next 3 periods. You also collected bimonthly data on electric usage charges.

Data: WaterCharges.MTW (available in the Sample Data folder).