This paper provides a demonstration of choosing the most appropriate statistical model and modeling techniques for large truck accident number prediction. The study is focused on examination of geometric parameters of two-lane rural segments of highways affecting single-vehicle large truck accidents. The study is done to specify geometric parameters of two-lane rural roads affecting large truck accidents where only one vehicle/combination vehicles is involved. The effects of controllable roadway design on single-vehicle large truck accident risk were quantitatively evaluated. Data from the Highway Safety Information System (HSIS) covering information about traffic accidents in Washington State for 2002-2005 is used for his study. The appropriateness of Negative Binomial, Zero Inflated Negative Binomial, Poisson and Zero Inflated Poisson regression models has been checked to create the reliable model for single-vehicle large truck accident prediction. The model-selection is based on possible value of the Vuong test overdispersion parameter (α) t-statistics. The ZINB regression model is chosen for modeling the expected frequency for single-vehicle large truck accidents. Created model provide a good statistical fit and provide reasonable results. The most significant geometric parameters influencing single-vehicle large truck accidents on two lane rural roads are average shoulder width and the percentage of roadway segment with curvature larger than 6 degrees. Traffic-related elements: annual average daily traffic (AADT), million vehicle miles traveled (MVMT) and truck percentage in the traffic stream on roadway segment have increasing effect on number of single-vehicle large truck accidents. Road segments of principal arterials have increasing effect too. It is found that MVMT is exclusive parameter for zero-accident state.