Page 89 - tmp_Neat

to the imbalance in the dependent variable. This inference was supported by the significant

skewness estimated in the scobit model. In addition, in the mixed logit model, uniform
distribution performed slightly better than a normal distribution, as suggested by Hensher and

Greene (2003) and Train (2003).

Overall, the results show similar signs and magnitudes for most of the estimated coefficients
across the three models. Nevertheless, there are some minor differences among the models.

Most of the variables are statistically significant in all models, except the number of lanes (5-

6 lanes), which is not significant in the scobit and mixed logit model, and the cross
intersection and road division marking (not divided) variables are not statistically significant

in the scobit model. Therefore, the discussion of the results in the rest of this section is based
on statistically-significant variables with similar signs and magnitudes of estimated

coefficients across the three models, especially for the mixed logit and skewed logit models,

since they performed better than the standard logit model.

72