A primary goal of the Ohio State Highway Patrol (OSHP) is to develop and
implement strategies for reducing the number of injury and fatal vehicle
crashes on Ohio highways. In order to help OSHP effectively allocate
its resources to reduce crash rates, the Statistical Consulting Service
(SCS) constructed a statistical model to forecast crash rates on all
interstates, US routes, and state routes throughout Ohio. Data for
the model were primarily drawn from Department of Public Safety crash
databases and GIS roadway files. Exploration of these data revealed
some consistent patterns that could be exploited for efficient modeling
in addition to some weaknesses in the data, such as miscoded and missing
data, that needed to be accounted for in modeling. Ultimately, a series
of ordinary Poisson regression models were fitted to the data using
resources from the Ohio Supercomputing Center. One final challenge of the
project was to develop a tool for presenting the massive model output in
a format that end users could readily understand. This challenge was
met by integrating model output with a dynamic visualization system,
Google Earth.
Meet the speaker in Room 212 Cockins Hall at 4:30
p.m. Refreshments will be served.