Measurement Error Models for Spatial Network Lattice Data: Analysis of Car Crashes in Leeds
Published in ArXiv, 2022
Recommended citation: Gilardi A., Borgoni R., Presicce L., Mateu J. (2022). "Measurement Error Models for Spatial Network Lattice Data: Analysis of Car Crashes in Leeds" arXiv:2201.02394. 1(1). https://arxiv.org/abs/2201.02394
Road casualties represent an alarming concern for modern societies demanding evidence-based interventions. Statistical models for road safety analysis typically include socio-economic variables and traffic volumes. However, the latter variables usually suffer from measurement error (ME), which can severely bias the statistical inference. This paper presents a Bayesian hierarchical model to analyse car crashes occurrences taking into account ME in the spatial covariates and the lattice structure of the road segments. Using a CAR prior, this work introduces a spatial dependence structure within the classical ME model. The suggested methodology is exemplified considering road collisions in the road network of Leeds (UK). Traffic volumes are approximated at the street segment level using an extensive dataset obtained from mobile devices. Estimation was carried out with the INLA methodology, which allows for computational advantages. Our results show that omitting ME adjustment considerably worsens the model’s fit and attenuates the effects of imprecise covariates.