Research Article

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

Table 5

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

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

0.50.48210.01793.57600.017527.4890
0.60.57340.02664.43070.021027.2577
0.50.56220.062212.43020.054010.4099
0.60.66580.065810.95830.063910.4166
0.30.30440.00441.45370.010329.4308
0.30.30600.00602.01600.03069.9954
−0.4−0.41010.01012.52800.0409−10.0344
1
0
0.60.59900.00100.16620.046112.9991
0
1
0
0.60.64770.04777.95000.10046.4540
1
0
1
−1−1.00680.00680.68160.0115−87.6890
−1−0.96710.03293.29030.0161−60.1612
−1−1.00760.00760.76470.0131−76.8664
−1−0.98730.01271.26820.0373−26.4510
0.30.31470.01474.91100.012525.2700
0.40.35090.049112.27750.03789.2838
0.50.49850.00150.29240.021023.7831
0.60.59060.00941.56250.07937.4485
1.51.51200.01200.80000.0081186.2896
1.51.47930.02071.38040.018779.1925
1.51.49930.00070.04560.0145103.4676
1.51.51560.01561.03930.053828.1629
0.50.45460.04549.07900.030414.9559
−1−1.04170.04174.17480.0476−21.8780
−0.8−0.87190.07198.99350.0445−19.5883
−1−1.01120.01121.12150.0158−63.9360
−0.8−0.81060.01051.31870.0127−63.6256
0.50.53790.03797.57200.023422.9583
0.20.18590.01417.04550.018410.0864
0.20.18770.01236.16000.018610.1024
0.50.48820.01182.36540.023221.0692
0.50.50940.00941.87680.021124.1120
0.20.19950.00050.26650.02049.7651
1
0
0
0
1
0.60.59520.00480.79270.052511.3353
0.60.63770.03776.28870.045713.9456
1.361.40020.04022.95340.107313.0519
0.60.69300.093015.49500.07768.9351
1.361.46280.10287.55510.100114.6143
Overall mean value0.02624.24000.0370

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