Richard Amoh-Gyimah presented his research on Modelling Crash Unobserved Heterogeneity Using Semi-Parametric Geographically Weighted Poisson Regression at the Australian Road Safety Conference (ARSC2016) held in Canberra, 6-8 September, 2016.
Abstract. Crash data are typically collected with reference to location dimension. Such data suffer from unobserved heterogeneity. The objective of this research is twofold: (1) to develop zonal crash prediction models using the Semi-Parametric Geographically Weighted Poisson Regression (S- GWPR) to address the issue of unobserved heterogeneity and (2) compare the performance of the S- GWPR with a non-spatial negative binomial (NB) model. The result indicates that by accounting for unobserved heterogeneity, the S-GWPR models performed better than the NB models. It was also found that unlike the NB models that show fixed parameters, all the variables except three have spatially varying coefficients in the S-GWPR.
0 Comments
Leave a Reply. |
AuthorDr. Meead Saberi, lecturer in transportation engineering, data guru, and urban scientist Archives
May 2019
Categories |