Random-Parameter Multivariate Negative Binomial Regression for Modeling Impacts of Contributing Factors on the Crash Frequency by Crash Types

Table 6

Predictive performance measure of two models (RPMNB and RPMP).

Crash type	RMSE		MAE		MAPE
	RPMNB	RPMP	RPMNB	RPMP	RPMNB	RPMP
Rear-end	7.413	8.374	6.551	6.941	1.433	1.638
Bumping-guardrail	11.070	10.833	9.781	9.551	3.639	2.919
Other crash	12.222	12.983	10.661	11.051	5.349	5.989
Noncasualty crash	1.643	1.597	1.253	1.252	0.136	0.128
Casualty crash	14.351	15.105	12.811	13.298	10.372	11.116
Total	46.699	48.892	41.055	42.092	20.929	21.789

Note: RPMNB = random-parameter multivariate negative binomial model. RPMP = random-parameter multivariate Poisson model. Measures where the RPMP model performs better.