Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
32 In this guide, recommended models by facility type were summarized. The univariate count model is recommended for rural two-lane highway segments and intersections. The multilevel model is recommended for rural multilane highway segments and intersections. The negative binomial ordered probit fractional split model is recommended for urban and suburban roads and intersections. Two separate steps may be needed to get predictions using the recommended models. If users do not have the driver and vehicle variables at the aggregated site level from other sources, the first step, which uses the crash data from the userâs jurisdiction to generate the QIE variables at the site level, should be performed, before the second step (crash prediction) can be run. Data requirements were discussed regarding running the QIE tool and performing the crash prediction step using the recommended models. Users need to ensure that the requirements are met before proceeding to the implementation tool. The implementation tool is a web-based R Shiny application that is available for download on the National Academies Press website (nap.nationalacademies.org) by searching for NCHRP Research Report 1047: Development and Application of Crash Severity Models for Highway Safety. The application helps users easily implement the recommended models without having to cal- culate the predictions manually. Additional error checking was added to the tool that indicates whether the input files match with each other. The results of the prediction models are automatically calibrated. The calibrated predictions should be appropriate for both planning-level and project-level purposes. Since the models developed in this project are fully specified models, no additional crash modification factors are needed. C H A P T E R 4 Summary