Transportation Research Record 1840 ■ 41 Paper No. 03-3883 A national-level safety analysis tool is needed to complement existing analytical tools for assessment of the safety impacts of roadway design alternatives. FHWA has sponsored the development of the Interactive Highway Safety Design Model (IHSDM), which is roadway design and redesign software that estimates the safety effects of alternative designs. Considering the importance of IHSDM in shaping the future of safety-related transportation investment decisions, FHWA justifi- ably sponsored research with the sole intent of independently validat- ing some of the statistical models and algorithms in IHSDM. Statistical model validation aims to accomplish many important tasks, including (a) assessment of the logical defensibility of proposed models, (b) assess- ment of the transferability of models over future time periods and across different geographic locations, and (c) identification of areas in which future model improvements should be made. These three activ- ities are reported for five proposed types of rural intersection crash prediction models. The internal validation of the model revealed that the crash models potentially suffer from omitted variables that affect safety, site selection and countermeasure selection bias, poorly mea- sured and surrogate variables, and misspecification of model func- tional forms. The external validation indicated the inability of models to perform on par with model estimation performance. Recommen- dations for improving the state of the practice from this research include the systematic conduct of carefully designed before-and-after studies, improvements in data standardization and collection practices, and the development of analytical methods to combine the results of before-and-after studies with cross-sectional studies in a meaningful and useful way. Effective safety management requires that engineers know the present safety performance of a roadway and how it will perform if contem- plated actions are taken. In effect, a reliable method for estimation of the safety performance of roadways under a host of potential future scenarios is needed. To this end, FHWA has sponsored research and development for a new approach that combines historical accident data, regression analysis, before-and-after studies, and expert judg- ment to make safety performance predictions that are expected to be better than those obtained by any of the individual approaches. A recent report documents an accident prediction algorithm for implementation of the new approach for two-lane rural highway sec- tions that includes road segments and five types of intersections (1). Ongoing efforts aim to produce similar documents for other types of facilities. A companion paper by Lyon et al. focuses on detailed aspects of this proposed algorithm (Lyon et al., pp. 78–86, this volume). This paper, in contrast, reports on the validation of individual crash models intended for use in the Interactive Highway Safety Design Model (IHSDM). Many articles in the literature report on the value and virtues of model validation (2–4). Widespread agreement on a precise definition of model validation is lacking in the research com- munity, although quite a bit has been written on the subject (2, 5). Val- idation can also be thought of as a requirement to demonstrate that a model is appropriate, meaningful, and useful for the purpose for which it is intended. Validation exercises are often associated with assessment of the prediction ability of a statistical model (2). However, it is possible for a model to predict an underlying data-generating process adequately but fail to illuminate and explain the nature of the underlying process. For this reason, model validation is broken down into two distinct aspects: internal validity and external valid- ity. Internal model validity, as applied in this research, is con- cerned with the ability of the intersection crash models to explain the underlying phenomenon, whereas external model validity is concerned with the ability of the models to predict crashes over time and space. The internal and external validation tools applied are described here. BACKGROUND AND METHODS Validation of Accident Prediction Models in IHSDM A statistical model, if internally valid, should agree with theoretical expectations in a number of respects. First, it should be consistent with established knowledge on the subject. Disagreement with past research should raise concern as to the plausibility of the results and require a greater burden of proof and explanation. Second, a good model should possess the salient and important features of the underlying system or phenomenon. Finally, the model or impor- tant parts of the model should agree with fundamental information and knowledge, such as physical mechanics, properties of ma- terials, and dynamics involved with crashes. It is the rigorous application of critical thinking about a phenomenon that can often raise doubt as to the validity of hypothesized relationships between variables and that serves to provide internal validation of statistical models. External validation is focused on the goodness of fit (GOF) of statistical models to independent data. Several GOF measures were used to assess model performance in this research. It is important to note that an objective assessment of the predictive Validation of FHWA Crash Models for Rural Intersections Lessons Learned Jutaek Oh, Craig Lyon, Simon Washington, Bhagwant Persaud, and Joe Bared J. Oh and S. Washington, Department of Civil Engineering, University of Arizona, Tucson, AZ 85721-0072. C. Lyon and B. Persaud, Department of Civil Engineering, Ryerson University, Toronto, Ontario M5B 2K3, Canada. J. Bared, Research and Development, FHWA, 6300 Georgetown Pike, McLean, VA 22101.