Multiple Imputation for Incomplete Traffic Accident Data Using Chained Equations

Missing value in traffic accident data prevents the discovery of the significant factors to reduce accident severity and even lead to an invalid conclusion. In previous studies, to handle this problem, researchers mainly tried to improve the methodologies to fit the incomplete data. In this paper, we propose a missing value imputation method. It can impute missing values in the traffic accident data set. The method is called multiple imputation by chained equations (MICE) which is flexible and practical. It can not only cope with univariate missing values but also multivariate missing values. The proposed algorithm is compared with two traditional imputation methods using two publicly available traffic accident datasets from New York. Furthermore, we test the performance of the model with different missing ratios. The imputations for continuous variables and discrete variables are analyzed separately. The results indicate that our proposed model outperforms the other two models under almost all situations.