A paper co-authored by UConn professor Dr. John Ivan of Civil & Environmental Engineering, Dominique Lord (Texas A&M) and Simon Washington (Berkeley), was the top-cited article published in the journal, Accident Analysis and Prevention during the last five years, with 68 citations.

The paper, entitled “Poisson, Poisson-gamma and Zero-inflated Regression Models of Motor Vehicle Crashes: Balancing Statistical Fit and Theory,” appeared in 2005, vol. 37, issue 1.

Dr. Ivan explained that it detailed the mechanistic aspects associated with motor vehicle crashes and evaluated the accuracy of various statistical models in predicting auto crashes. He said that many researchers analyze crash figures using models that classify transportation states as either crash-free or as following a crash distribution.

The article summarized the authors’ study of simulated experiments, which led them to conclude that zero-crash observations arise in situations characterized by low traffic volume, very short-time observation intervals and low crash occurrence rates.

“We concluded that it is more important to define a prediction framework that can functionally explain the phenomena than to choose one that provides a good statistical fit for the data.  Ultimately, the goal of any statistical model is to generate knowledge that is transferable to other contexts,” explained Dr. Ivan.