9 February 1988

Example 6:  Use of modelspan to facilitate finding an acceptable
            model for an economist's index series.  We could not 
            find a model with good diagnostics for the full series.
            A graph of the log-transformed series suggests it has a 
            different growth rate after the first three years.  An 
            acceptable model was found for the data span beginning 
            January, 1989.  Note the improvements to all the 
            modeling diagnostics.



# Example 6:  index0.spc  

# An economist's index series.  Difficult to find an acceptable model without
# a modelspan statement.  

Series{
  title = "Index Series Run with Full-Series Modeling"
  file="index.dat"
  start = 1986.jan
}
transform{  function=log  }
regression{ variables=( AO1991.Apr )        }
arima{   
         model=([1,6] 1 0)(0 1 1)
}
estimate{   }
check{print=all}
forecast{  maxlead=36   }



# Example 6: index.spc  

# Modelspan allows us to find an acceptable model for an economist's index series.

Series{
  title = "Index Series Run using Modelspan"
  file="index.dat"
  start = 1986.jan
  modelspan = (1989.jan,)
}
transform{  function=log  }
regression{ variables=( AO1991.Apr )        }
arima{   
         model=([1,6] 1 0)(0 1 1)
}
estimate{   }
check{print=all}
forecast{  maxlead=36   }