County-level commodity flow data were commercially procured to describe freight flows into, out of, within, and through Virginia. With the use of these data, Virginia's key commodities were identified and their flows were assigned to county-level origin-destination tables. Predictive equations of freight-generation and -attraction relationships for each of Virginia's 15 key commodities were developed. Explanatory variables were investigated and defined, including sources. A strategy for developing regression equations using a series of weighted regressions was developed using a series of robust and stepwise regressions to minimize the effects of outliers. For each key commodity with a two-digit Standard Transportation Commodity Classification code, a set of generation and attraction equations was developed, including relationships for nonoutliers, first-order outliers, and second-order outliers. In addition, several socioeconomic variables were identified that significantly affect freight generation and attraction within Virginia. It was concluded that robust regression is an appropriate tool for modeling freight generation, especially when outliers are present in the data set.
Abstract