Development of Integrated Choice and Latent Variable (ICLV) Models Using Matrix-Based Analytic Approximation and Automatic Differentiation Methods on TensorFlow Platform
Table 6
Simulation results for the four-alternative ICLV model with the TVBS method.
Parm.
True value
Estimate
Abs. bias
APB (%)
Std. err.
t-test
value
0.5
0.4820
0.0180
3.6006
0.0175
27.4956
0.6
0.5733
0.0267
4.4550
0.0210
27.2643
0.5
0.5621
0.0621
12.4248
0.0541
10.3956
0.6
0.6657
0.0657
10.9532
0.0640
10.4023
0.3
0.3042
0.0042
1.3913
0.0104
29.3686
0.3
0.3058
0.0058
1.9333
0.0305
10.0157
−0.4
−0.4098
0.0098
2.4453
0.0408
−10.0548
1
—
—
—
—
—
—
0
—
—
—
—
—
—
0.6
0.5986
0.0014
0.2263
0.0460
13.0038
0
—
—
—
—
—
—
1
—
—
—
—
—
—
0
—
—
—
—
—
—
0.6
0.6472
0.0472
7.8628
0.1002
6.4570
1
—
—
—
—
—
—
0
—
—
—
—
—
—
1
—
—
—
—
—
—
−1
−1.0068
0.0068
0.6844
0.0115
−87.6833
−1
−0.9671
0.0329
3.2898
0.0161
−60.1169
−1
−1.0078
0.0078
0.7807
0.0131
−76.7399
−1
−0.9876
0.0124
1.2408
0.0373
−26.4477
0.3
0.3148
0.0148
4.9407
0.0125
25.2747
0.4
0.3509
0.0491
12.2728
0.0379
9.2709
0.5
0.4989
0.0011
0.2154
0.0210
23.7248
0.6
0.5913
0.0087
1.4555
0.0793
7.4595
1.5
1.5120
0.0120
0.8026
0.0081
186.2650
1.5
1.4793
0.0207
1.3799
0.0187
79.0895
1.5
1.4996
0.0004
0.0299
0.0145
103.1776
1.5
1.5160
0.0160
1.0677
0.0538
28.1597
0.5
0.4544
0.0456
9.1144
0.0304
14.9438
−1
−1.0425
0.0425
4.2479
0.0476
−21.8930
−0.8
−0.8728
0.0728
9.1025
0.0445
−19.6065
−1
−1.0116
0.0116
1.1607
0.0158
−63.9212
−0.8
−0.8109
0.0109
1.3567
0.0127
−63.6051
0.5
0.5382
0.0382
7.6404
0.0234
22.9612
0.2
0.1860
0.0140
6.9835
0.0185
10.0749
0.2
0.1878
0.0122
6.1180
0.0186
10.0900
0.5
0.4887
0.0113
2.2598
0.0232
21.0338
0.5
0.5099
0.0099
1.9814
0.0212
24.0589
0.2
0.1997
0.0003
0.1510
0.0204
9.7806
1
—
—
—
—
—
—
0
—
—
—
—
—
—
0
—
—
—
—
—
—
0
—
—
—
—
—
—
1
—
—
—
—
—
—
0.6
0.5923
0.0077
1.2832
0.0524
11.3058
0.6
0.6382
0.0382
6.3702
0.0458
13.9236
1.36
1.3895
0.0295
2.1659
0.1074
12.9430
0.6
0.6875
0.0875
14.5838
0.0775
8.8672
1.36
1.4629
0.1029
7.5643
0.1001
14.6142
Overall mean value
0.0259
4.2037
0.0370
Note. “—” indicates that the corresponding parameter is fixed at the true value.