Research Article

Development of Integrated Choice and Latent Variable (ICLV) Models Using Matrix-Based Analytic Approximation and Automatic Differentiation Methods on TensorFlow Platform

Table 4

Simulation results for the four-alternative ICLV model with the ME method.

Parm.True valueEstimateAbs. biasAPB (%)Std. err.t-test value

0.50.48210.01793.58060.017527.4871
0.60.57340.02664.43450.021027.2569
0.50.56230.062312.45820.054510.3235
0.60.66590.065910.98580.064510.3299
0.30.30390.00391.28330.010329.4528
0.30.30590.00591.96930.030610.0031
−0.4−0.40990.00992.48170.0408−10.0422
1
0
0.60.59820.00180.30820.046013.0071
0
1
0
0.60.64750.04757.92420.10076.4324
1
0
1
−1−1.00680.00680.68180.0115−87.6878
−1−0.96710.03293.29330.0161−59.9248
−1−1.00810.00810.80700.0131−76.7456
−1−0.98750.01251.25310.0373−26.4410
0.30.31470.01474.91600.012525.2686
0.40.35080.049212.30200.03819.2070
0.50.49960.00040.08280.021023.7820
0.60.59100.00901.50230.07937.4520
1.51.51200.01200.80040.0081186.2783
1.51.47920.02081.38350.018878.6552
1.51.49990.00010.00340.0145103.2049
1.51.51580.01581.05530.053828.1516
0.50.45470.04539.05600.030414.9540
−1−1.04160.04164.16160.0476−21.8600
−0.8−0.87280.07289.10370.0445−19.5951
−1−1.01150.01151.14900.0158−63.8326
−0.8−0.81070.01071.34240.0128−63.5246
0.50.53800.03807.60640.023422.9537
0.20.18580.01427.07700.018610.0106
0.20.18770.01236.14750.018710.0234
0.50.48910.01092.17560.023221.0530
0.50.51030.01032.06320.021224.0984
0.20.19970.00030.15650.02049.7718
1
0
0
0
1
0.60.59560.00440.73650.052611.3276
0.60.64070.04076.77720.046013.9313
1.361.40080.04082.99990.107413.0425
0.60.69390.093915.64220.07798.9014
1.361.47230.11238.25420.100714.6142
Overall mean value0.02664.26910.0371

Note: “—” indicates that the corresponding parameter is fixed at the true value.