
194 ◾ Transportation Statistics and Microsimulation
as x increases, then the assumption of constant variance is probably
not appropriate. When a nonconstant variance is suspected, the data
are typically handled by either performing weighted least squares or by
transformations, such as Box-Cox transformations. An excellent refer-
ence for handling nonconstant variance is Carroll and Rupert (1988).
Note that if the assumption is violated, this is usually not considered
a fatal error. is is because the estimated slope and intercept are still
unbiased, but unnecessarily inaccurate. e resulting population vari-
ances would be bigge ...