User-Specified Trend Models in Statistical Forecasting

Overview

In some situations statistical forecasting based on historical data is inappropriate e.g. for new products with little or no history, or for products which are about to enter totally new markets. In such situations, it may still be helpful to project future trends by either using selected mathematical trend curves with user specified parameters. 

For each of these models, you may:

• Specify the nature of the trend curve
• Specify the seasonal factors to be used 
• IFP will then produce graphical and statistical analyses of the forecasts generated by the selected options.
• You are then given the option of modifying the vertical position and/or slope of their trend curve using the graph modification procedure
• Any user specified trend curve may start from any period after the latest actuals month

Model options

User-Specified Linear Trend

• The initial value is the value of the underlying trend at the selected start period
• The slope of the trend is measured in terms of units per month
• The initial value of the trend must be greater than zero but the slope may be positive or negative

User-Specified Exponential Trend

• The initial value of the trend must be greater than zero
• The (annual) % growth rate may be positive or negative

User-Specified Modified Exponential

• The saturation level is the maximum monthly value for the trend
• The growth parameter measures the rate at which the underlying trend approaches the saturation level
• The initial value of the trend must be greater than zero
• The saturation level must exceed the initial trend value by at least 10 %.
• The growth parameter should normally be in the range 0.70 to 0.99, the example below using 0.95
• The lower the growth parameter, the faster the trend approaches the saturation level

User-Specified S-Curve

• The saturation level is the maximum monthly value for the trend
• The growth parameter measures the rate at which the underlying trend approaches the saturation level
• The initial value of the trend must be greater than zero
• The saturation level must exceed the initial trend value by at least 10 %.
• The growth parameter should normally be in the range 0.70 to 0.99, th example below using 0.92
• The lower the growth parameter, the faster the trend approaches the saturation level