Page 55 - tmp_Neat

In this study, crashes involving no injury were selected as the reference category for the
dependent variable. Therefore, the estimated coefficients show the impacts of the

contributing factors on fatal, severe and minor injury relative to the reference category (no
injury). Although the traditional 95% level of confidence was used to select variables, some

insignificant variables were retained in the model as long as it was statistically significant for
at least one of the injury outcomes. This was done to facilitate the interpretation of the results

(Kockelman and Kweon, 2002; Tay et al., 2008; Tay et al., 2009; Tay et al., 2011). In

addition, the marginal effects of each independent variable were calculated to facilitate the
interpretation of the results. The marginal effects provided estimates of the changes in the

probabilities of the different injury outcomes due to a unit change in the independent

variables.

4.2.3 Random Parameter Logit (Mixed Logit Model)

A summary of existing studies on the random parameter logit model is presented in Section
2.2.1.5 of Chapter 2. Random parameters or mixed logit is applied to allow the possibility

that the parameters may vary across observations (Washington et al., 2010). Some
researchers have chosen to use the random coefficient logit or probit model to allow for

heterogeneous effects and correlations in unobserved factors (Milton et al., 2008; Kim et al.,

2010; Anastasopoulos & Mannering, 2011; Tay, 2015). Random parameter models,
especially the random parameter logit or mixed logit models, have increasingly been used in

traffic safety studies to analyse both crash frequency and severity (Lord & Mannering 2010;
Savolainen et al. 2011). To develop the mixed logit model, this study follows Milton et al.,

(2008) and starts with the severity function as below:

(4.3)

where,

= is a linear function for determining the injury severity category i to road user n,

= a vector of estimated coefficients,

= a vector of explanatory variables,

= is an error term.

44