Previous Topic: 14.3.2.5 Regression Model Output GraphNext Topic: 14.3.2.7 Forecasting Check Report


14.3.2.6 Regression Model Residuals Graph


This graph displays residual values from the regression model
operation.  Residual values are differences observed between
the actual and predicted values of the dependent (Forecast)
variable.  The linear regression statistics are displayed on
each page of the report for reference.  This graph can be
used to help evaluate the success of the model by ensuring
that the model is functioning properly.

REGRESSION MODEL RESIDUALS GRAPH 1 PLOT RESIDUAL VS ACTUAL VALUES

MODEL STATISTICS: M= 101.764500 B= -1447836 R2= 0.1694 -+-------------+-------------+-------------+-------------+-------------+-------------+-------------+-------------+- | | | | 26 + + | | | | | | 21 + + | | | | | | 16 + + | | M | | o | | d 11 + + e | | l | | | R | R 5 + R R + e | RR R R R | s | RR RR R R R R | i | RR R R R R R | d 0 +------------------------------------------------------------------------R------R-----R---------------------------+ u | RR R R RR | a | R | l | R R | 5 + + V | R | a | | l | | u 11 + R R + e | | | | | | 16 + + | | | | | | 21 + + | | | | | | 26 + R + | | -+-------------+-------------+-------------+-------------+-------------+-------------+-------------+-------------+- 0 6 11 17 22 28 33 39 44 Adjusted CPU Time NOTE: 5 obs hidden. LEGEND: RESIDUAL VALUES - 'R'


 Figure 14-8.  Regression Model Residuals Graph

Anytime a linear regression model is used, the residual
values should be plotted against the actual values of the
dependent (Forecast) variable and examined for statistical
abnormalities.  The residual values should cluster around
zero and should be present in a somewhat "random" pattern.
Residuals that track the original data values or form other
specific patterns can indicate problems within the model.
Consult any basic statistics book for additional details.