Transportation Statistics and Microsimulation

220    ◾    Transportation Statistics and Microsimulation

Poisson regression model corresponds to the negative binominal regres-

sion model when the value of k is 0.

Example 11.3

Recall that the crash data in Example 11.1 revealed an overdispersion problem.

Although an overdispersed Poisson model can be used to correct the esti-

mated covariance matrix when overdispersion is modest, an negative binomial

regression model is, in general, a more preferred model when the data are

overdispersed. Table 11.5 contains the SAS output for the negative binominal

regression obtained by using SAS PROC GENMOD as follows:

PROC GENMOD DATA=RSBS_2L_NM_LS;

MODEL Total_Crashes_for_3_years = Right_Lane Right_Shoulder

/ DIST=NB LINK=Log OFFSET=LogEXPO3;

RUN; ...