6. SIMPLE REGRESSION ANALYSIS › 6.2 Usage Guidelines

6.2 Usage Guidelines

The main factor you must assess for Simple Regression
Analysis is the number of future intervals that you wish to
estimate using the linear regression model.  As with Profile
and Trending Analysis, discussed in Chapter 5, the number of
intervals that you forecast based on an historical data
series is dependent on two characteristics of the historical
data series:

    o  The length (that is, the number of observations) of
       the historical data series

    o  The behavior of the historical data series

The number of forecast intervals that you specify should
never exceed the length of the historical data series used to
produce the model.  In fact, the probable error associated
with each future value increases as a function of the number
of the future values estimated.  Quantitative techniques for
determining these confidence limits for regression-based
forecasts are discussed in Chapter 7, Univariate Model
Forecasting.  Since the probability of error increases with
forecast length, we recommend a maximum forecast length of
one-third to one-half the length of the historical data
series.

You must also consider the qualitative behavior of the
historical data series when you select the number of
intervals to be forecast.  If the differences between the
regression line and the historical data series are small,
then you can specify a longer forecast.  If the differences
are large (that is, greater than 10 to 15 percent of the
actual value), then you should specify a very short period.