Vol 10 Issue 1 January 2023-February 2023
Azad Abdulhafedh
Abstract: This paper addresses the Zero-inflated Poisson (ZIP) regression model as an effective way to handle the excess zeros that usually exist in vehicle crash data and to allow for possible overdispersion in the data. The ZIP model is based on a zero-inflated probability distribution, that allows for frequent zero-valued observations. When the number of zeros is large that the data do not fit standard distributions (e.g., normal, Poisson, binomial, negative-binomial, and beta), the data is referred to as zero inflated. A dual state crash system is assumed in the ZIP model, in which one state is the zero crash state that can be regarded as virtually safe during the observation period, while the other state is the non-zero crash state. This paper starts by applying a multiple linear regression model, a Poisson regression model, a Negative Binomial regression model and then introduces the ZIP model to analyze the 2013-2015 crash data for the Interstate I-94 in the State of Minnesota in the US. Results show that the ZIP model overperformed the other models by fitting the crash data much better and was able to capture almost all the independent variables in the model and make them statistically significant in the analysis after being insignificant by the other models.
Keywords: Zero-Inflated Poisson Regression, ZIP model, Crash Frequency, Multiple Linear Regression, Poisson Regression, Negative Binomial Regression.
Title: Incorporating Zero-Inflated Poisson (ZIP) Regression Model in Crash Frequency Analysis
Author: Azad Abdulhafedh
International Journal of Novel Research in Interdisciplinary Studies
ISSN 2394-9716
Vol. 10, Issue 1, January 2023 - February 2023
Page No: 6-18
Novelty Journals
Website: www.noveltyjournals.com
Published Date: 11-February-2022
