Mixed Multinomial Probit Model Accommodating Flexible Covariance Structure and Random Taste Variation: An Application to Commute Mode Choice Behavior

Table 1

Simulation results for a four-alternative MMNP model of 500,000 observations when random coefficients are associated with generic variables.

Parameters	True values	Estimates	Std. err.	t-stat
	−0.5	−0.498	0.074	−6.776^∗∗∗
	−0.4	−0.377	0.067	−5.640^∗∗∗
	−0.3	−0.259	0.047	−5.551^∗∗∗
	0.5	0.511	0.044	11.732^∗∗∗
	0.3	0.306	0.027	11.563^∗∗∗
	0.2	0.202	0.015	13.503^∗∗∗
	0.1	0.099	0.007	13.431^∗∗∗
	0.2	0.206	0.018	11.658^∗∗∗
	0.4	0.404	0.035	11.722^∗∗∗
	0.6	0.606	0.051	11.833^∗∗∗
	0.5	0.528	0.045	11.667^∗∗∗
	0.2	0.203	0.016	12.396^∗∗∗
	0.4	0.379	0.015	25.196^∗∗∗
	0.5	0.541	0.013	40.782^∗∗∗
	0.6	0.608	0.008	74.205^∗∗∗
	0.3	0.299	0.011	28.064^∗∗∗
	0.4	0.401	0.010	42.107^∗∗∗
	1.0	1.000	—	—
	0.5	0.511	0.087	5.843^∗∗∗
	−0.3	−0.341	0.105	−3.250^∗∗∗
	0.6	0.524	0.080	6.521^∗∗∗
	1.0	1.000	—	—
	−0.4	−0.372	0.127	−2.932^∗∗∗
	0.5	0.532	0.096	5.517^∗∗∗
	1.0	1.000	—	—
	0.3	0.300	—	—
	1.0	1.000	—	—

Note. “—” the corresponding parameter is fixed; ^∗∗∗significant at 0.01 level.