Page 192 - Planning and Design of Airports

For ecasting for Airport Planning 159

There are a great variety of techniques which are used in econo-
metric modeling for airport planning. Classical trip generation and
gravity models are quite common in forecasting passenger and air-
craft traffic. Simple and multiple regression analysis techniques, both
linear and nonlinear, are often applied to a great variety of forecasting
problems to ascertain the relationships between the dependent vari-
ables and such explanatory variables as economic and population
growth, market factors, travel impedance, and intermodal competi-
tion. The form of the equations used in multiple linear regression
analysis is given in Eq. (5-1).

Y = a + a X + a X + a X + ··· + a X (5-1)
est 0 1 1 2 2 3 3 n n
where Y = dependent variable or variable which is being
est
forecast
X , X , X ,…, X = dependent variables or variables being used to
1 2 3 n
explain the variation in the dependent variable
a , a , a , a ,…, a = regression coefficients or constants used to cali-
0 1 2 3 n
brate the equation
There are many statistical tests which can be performed to deter-
mine the validity of econometric models in accurately portraying
historical phenomena and to reliably project demand. Though the
constants may be found to define the general equation of the model,
it is possible that the range of error associated with the equation may
be large or that the explanatory variables chosen do not directly
determine the variation in the dependent variable.
There may be a tendency when performing sophisticated mathe-
matical modeling to become disassociated from the significance of
the results. It is incumbent upon the analyst to consider the reason-
ableness as well as the statistical significance of the model. Adequate
consideration must be given to the rationality of the functional form
and variables chosen for the analysis, and to the logic associated with
calibrated constants.
In many cases it is essential to determine the sensitivity of fore-
casts to changes in the explanatory variables. If a particular design
parameter being forecast varies considerably with a change in a
dependent variable, and there is a significant degree of unreliability
in this independent variable, then a great deal of confidence cannot
be placed upon the forecast and, more importantly, the design based
upon the forecast. Tests are usually performed to determine the
explanatory power of the independent variables and their interrela-
tionships. The analyst should carefully investigate the sensitivity of
projections within the expected variation of explanatory variables. It
is also possible that certain explanatory variables do not significantly
affect the modeling equation and the need for collecting the data
associated with these variables required for projections could be