A Fast and Efficient Estimation of the Parameters of a Model of Accident Frequencies via an MM Algorithm

Table 1

Results for Scenario 1 ( and ). Values in brackets are standard deviations.

		MM	SqS3	NR	BFGS
		0.826 (0.176)	0.826 (0.176)	0.826 (0.176)	0.824 (0.174)
		0.649 (0.068)	0.649 (0.068)	0.649 (0.068)	0.648 (0.068)
		0.351 (0.068)	0.351 (0.068)	0.351 (0.068)	0.352 (0.068)
		0.247 (0.062)	0.247 (0.062)	0.247 (0.062)	0.246 (0.061)
		0.753 (0.062)	0.753 (0.062)	0.753 (0.062)	0.754 (0.061)
	Convergence proportion (%)	100	100	99.5	76.1
	Iterations	24 (4.1)	7 (1)	5.5 (1.5)	13 (0.1)
	CPU time (secs)	0.004	0.002	0.002	0.087
	Time ratio	1.00	0.51	0.62	23.63
	Log-likelihood	-126.66	-126.66	-126.67	-126.71
	MSE
		0.800 (0.016)	0.800 (0.016)	0.800 (0.016)	0.800 (0.030)
		0.650 (0.007)	0.650 (0.007)	0.650 (0.007)	0.650 (0.010)
		0.350 (0.007)	0.350 (0.007)	0.350 (0.007)	0.350 (0.010)
		0.250 (0.006)	0.250 (0.006)	0.250 (0.006)	0.250 (0.009)
		0.750 (0.006)	0.750 (0.006)	0.750 (0.006)	0.750 (0.009)
	Convergence proportion (%)	100	100	99.7	77.3
	Iterations	31.2 (4.3)	8.2 (1)	5.7 (1.4)	16 (0.5)
	CPU time (secs)	0.006	0.003	0.003	0.165
	Time ratio	1.00	0.45	0.50	26.25
	Log-likelihood	-12832.34	-12832.34	-12832.39	-12838.15
	MSE