Weighted Estimation of Trends in Statistical Forecasting

IFP includes an option to use weighted estimation of the path of the trends. Here, more weight is put on recent data than on older data.

Weighted estimation is useful where the path of the underlying trend has changed recently. The weighting of historical data is achieved by using an appropriate discount factor in the range 50% to 99%.

Weighted estimation is available for the following types of trend:

• average
• linear trend
• exponential trend

Note that weighting does not change the shape of the trend itself.

Discount Factor

This determines the rate at which IFP effectively discounts the value of historical data, according to its 'age'.  For example, a discount factor of, say, 90% would discount historical demand data at a rate of 90% per month; i.e. IFP would only put 90% of the weight on the demand for, say, March 2011 as it would on that for April 2011.  A discount factor of 90% would produce weights that decline in the pattern 100, 90, 81, 72..... .

The lower the discount factor the more emphasis is put on recent demand data when estimating the future path of the trend.