Loading…
Selected models for correlated traffic accident count data
Accident counts are correlated in many ways that induced by unexplained heterogeneity. The sources of the unexplained of heterogeneity were identified as spatial, temporal and categorical. Multiple models were introduced to cater one or two of these factors. However, to date, there is no single mode...
Saved in:
Main Author: | |
---|---|
Format: | Conference Proceeding |
Language: | English |
Subjects: | |
Online Access: | Get full text |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Summary: | Accident counts are correlated in many ways that induced by unexplained heterogeneity. The sources of the unexplained of heterogeneity were identified as spatial, temporal and categorical. Multiple models were introduced to cater one or two of these factors. However, to date, there is no single model incorporates these three factors together. Hence, we review three selected models for correlated count data; the Generalized ARMA (GLARMA) model, model with lagged observation and seemingly unrelated negative binomial model (SUNB). The aim of this paper is to conduct an initial study to assess the stability of the selected models. Through simulation study, the strength and the weakness of these models are also identified to evaluate their potential for application to correlated accident count data. Based on the simulation results, the GLARMA model is found to give inconsistent estimates and need to have an adequate sample of data to provide a significant result. Models with lagged observation provide satisfactory results since the temporal structure is inserted as fixed effects. Meanwhile, the SUNB model needs to have a fixed variance component to ensure the stability of the parameter estimates. Overall, the simulation study conducted reveals that model with fixed components structure provide more stable estimates compared to models with random effects. |
---|---|
ISSN: | 0094-243X 1551-7616 |
DOI: | 10.1063/1.4954629 |