2. Spatial Autocorrelation
The result of Chi-square indicates variables including speeding, road location and times of a day are statistically significantly relevant to the traffic fatality. While other variables such as road types, week and season were not influence on the severity. However, Chi-square does not help to examine whether these correlations were positive or negative, and it also cannot tell the spatial relationships. Thus, the Moran’s I is necessary to investigate the spatial autocorrelation among the fatal traffic accidents with each significant features.
ArcGIS Global Moran’s I tool evaluates whether the pattern of events is clustered, dispersed or random, by giving a set of features and an associated attribute. The output file computes the Index value, Z value, and P value. The Index value calculates the expected value and observed value of a random distribution. It usually ranges from negative 1 to positive 1. When the index value is positive, it indicates a positive spatial autocorrelation; when the index is negative, it reflects a negative spatial autocorrelation. When the index is 0, it means a random distribution of accidents. The index value cannot be interpreted unless with the Z and P value. Only when the P value us significant (smaller than 0.05), the Moran’s I index is meaningful.
​
Table 1. Summary results of the Global Moran's I tool showing the Moran's I Index, Z score, and P value for fatal accidents at intersections, involving speeding, during morning rush hours, and during evening rush hours
By setting different significant variables in the Moran's I spatial autocorrelation, we ran the tool four times. The result indicated all the 4 variables have positive spatial autocorrelation since the Moran’s indexes were near 1, z values were positive and the p values were almost 0 (Table 1). The fatal accidents with speeding variable had an unusual Moran index of 1.3 (Tables 1). One possible explanation is due to the data speeding accidents were too skewed, that is not a normal distribution.