--- category: "[[Clippings]]" author: "[[I. Hilakivi]]" title: "Using logistic regression to estimate the influence of accident factors on accident severity" source: https://www.sciencedirect.com/science/article/abs/pii/S0001457501000732 clipped: 2024-07-21 published: topics: tags: [clippings] --- ## Introduction Accident severity is of special concern to researchers in traffic safety since this research is aimed not only at prevention of accidents but also at reduction of their severity. One way to accomplish the latter is to identify the most probable factors that affect accident severity. This study aims at examining not all factors, but some believed to have a higher potential for serious injury or death, such as accident location, type, and time; collision type; and age and nationality of the driver at fault, his license status, and vehicle type. Other factors were not examined because of substantial limitations in the data obtained from accident reports. Logistic regression was used in this study to estimate the effect of the statistically significant factors on accident severity. Logistic regression and other related categorical-data regression methods have often been used to assess risk factors for various diseases. However, logistic regression has been used as well in transportation studies. A brief literature review follows of the use of this type of regression in traffic safety research. ## Section snippets ## Literature review Regression methods have become an integral component of any data analysis concerned with the relationship between a response variable and one or more explanatory variables. The most common regression method is conventional regression analysis (CRA), either linear or nonlinear, when the response variable is continuous (iid). However, when the outcome (the response variable) is discrete, CRA is not appropriate. Among several reasons, the following two are the most significant: - 1 The response ## Theoretical background of logistic regression It is important to understand that the goal of an analysis using logistic regression is the same as that of any model-building technique used in statistics: to find the best fit and the most parsimonious one. What distinguishes a logistic regression model from a linear regression model is the response variable. In the logistic regression model, the response variable is binary or dichotomous. The difference between logistic and linear regression is reflected both in the choice of a parametric ## Model description The dependent variable in this research, *ACCIDENT,* is of the dichotomous type and stands for accident severity. It should be mentioned that the definition of injury in this study does not overlap with the definition of fatality since the first includes those who were involved in accidents and left the hospital within 6 months after treatment. Each accident in the sampled data was categorized as either non-fatal or fatal. The logistic model used isP(non\-fatalaccident)=π(x)=eg(x)1+eg(x)and thusP( ## Data description The data set used in this study was derived from a sample of 560 subjects involved in serious accidents reported in traffic police records in Riyadh, the capital of Saudi Arabia. Only accidents occurring on urban roads in Riyadh were examined. Unfortunately, police reports at accident sites do not describe injuries in much detail because of the lack of police qualifications and training as well as facilities needed to perform complex examinations. Also, medical reports are hard to obtain ## Reduction of design variables As can be seen from Table 1, some of the categorical variables have several levels, so several design variables are needed for each. Generally speaking, it is more convenient to have as few design variables as possible in order to simplify the model interpretation. In other words, the more design variables the model includes, the more difficult the interpretation becomes. Thus, an attempt was made in the early stages of this study to reduce the number of design variables. However, care is ## Development of logistic model The backward selection process of logistic regression was followed. First, all the variables with no interactions (referred to here as the saturated model; Fig. 3) were tested on the basis of the deviance and the Wald (*W*) statistic as defined in , , respectively. The goal was to eliminate, at the beginning, those variables that were not significant and then continue with testing interaction effects with only significant variables. Table 5 presents the results from fitting all the explanatory ## Logit model According to the previous analysis, the logit model with the significant variables is as follows:ĝ(x)=−2.029+0.9697LOC(2)−0.3558CAUS(2)+0.2130CAUS(3)−0.8971CAUS(4)−0.6705CAUS(5)Hence the logistic regression model developed in this study isπ(x)=e−2.029+0.9697LOC(2)−0.3558CAUS(2)+0.2130CAUS(3)−0.8971CAUS(4)−0.6705CAUS(5)1+e−2.029+0.9697LOC(2)−0.3558CAUS(2)+0.2130CAUS(3)−0.8971CAUS(4)−0.6705CAUS(5) Once the model has been fit, the process of assessment of the model begins. Several tests, including ## Model interpretation Interpretation of any fitted model requires the ability to draw practical inferences from the estimated coefficients. The estimated coefficients for the independent variables represent the slope or rate of change of the dependent variable per unit of change in the independent variable. Thus, interpretation involves two issues: determining the functional relationship between the dependent variable and the independent variable (i.e. the link function; McCullagh and Nelder, 1982) and appropriately ## Conclusions Since the response variable is of a binary nature (i.e. has two categories — fatal or non-fatal), the logistic regression technique was used to develop the model in this study. The intent was to provide a demonstration of a model that can be used to assess the most important factors contributing to the severity of traffic accidents in Riyadh. On the basis of traffic police accident data, nine explanatory variables were used in the model development process. Using the concept of deviance together Copyright © 2002 Elsevier Science Ltd. All rights reserved.