A Comprehensive Analysis for the Heterogeneous Effects on Driver Injury Severity in Single-Vehicle Passenger Car and SUV Rollover Crashes

Table 3

Summary statistics and the likelihood test results for six models varying datasets.

Dataset	Passenger car		SUV		Full dataset
	TOL	RPOL	TOL	RPOL	TOL	RPOL
Number of observations,	2400	2400	2360	2360	4760	4760
Number of parameters,	17	22	16	24	17	22
Log likelihood at zero,	−2991.7	−2991.7	−2937.1	−2937.1	−6168.4	−6168.4
Log likelihood at convergence,	−2747.4	−2730.4	−2701.7	−2681.9	−5669.0	−5651.8
AIC	5528.8	5504.8	5435.4	5411.7	11372.0	11347.6
Likelihood-ratio tests for the superiority of the RPOL model
	34		39.6
Degrees of freedom	5		8
-value	<0.005		<0.005
Likelihood-ratio tests for the necessity of developing passenger car and SUV models separately
					439.8	479
Degrees of freedom					16	24
-value					<0.005	<0.005

Note: TOL represents traditional ordered logit model, while RPOL represents random parameters ordered logit model; All TOL models have a constant.