Definition

Holt's linear exponential smoothing method is similar in principle to Brown's except that it smooths data separately.

X = actual value
F = the forecast value
a = alpha value
t = time period of integer
m = time period ahead

Potential Flaws

1. This is a weighted average where two critical assumptions are being made. The first assumption is that all of the past should be used and the second assumption is the weights that each historical value will have are known and recursive in form and are computable from two estimated and changing constants , a and b.

2. There is no provision for dealing with Pulses, Seasonal Pulses, Level Shifts or Local Time Trends.

3. There is no provision for changing variance, changing model or changing parameters.

4. There is an assumption of orthogonality or independence between the two model coefficients. There is no provision for interdependence.

4. There is an assumption about the degree and form of change that effects the two coefficients.