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 value
Estimate
Abs. bias
APB (%)
Std. err.
t-test
value
0.5
0.4821
0.0179
3.5806
0.0175
27.4871
0.6
0.5734
0.0266
4.4345
0.0210
27.2569
0.5
0.5623
0.0623
12.4582
0.0545
10.3235
0.6
0.6659
0.0659
10.9858
0.0645
10.3299
0.3
0.3039
0.0039
1.2833
0.0103
29.4528
0.3
0.3059
0.0059
1.9693
0.0306
10.0031
−0.4
−0.4099
0.0099
2.4817
0.0408
−10.0422
1
—
—
—
—
—
—
0
—
—
—
—
—
—
0.6
0.5982
0.0018
0.3082
0.0460
13.0071
0
—
—
—
—
—
—
1
—
—
—
—
—
—
0
—
—
—
—
—
—
0.6
0.6475
0.0475
7.9242
0.1007
6.4324
1
—
—
—
—
—
—
0
—
—
—
—
—
—
1
—
—
—
—
—
—
−1
−1.0068
0.0068
0.6818
0.0115
−87.6878
−1
−0.9671
0.0329
3.2933
0.0161
−59.9248
−1
−1.0081
0.0081
0.8070
0.0131
−76.7456
−1
−0.9875
0.0125
1.2531
0.0373
−26.4410
0.3
0.3147
0.0147
4.9160
0.0125
25.2686
0.4
0.3508
0.0492
12.3020
0.0381
9.2070
0.5
0.4996
0.0004
0.0828
0.0210
23.7820
0.6
0.5910
0.0090
1.5023
0.0793
7.4520
1.5
1.5120
0.0120
0.8004
0.0081
186.2783
1.5
1.4792
0.0208
1.3835
0.0188
78.6552
1.5
1.4999
0.0001
0.0034
0.0145
103.2049
1.5
1.5158
0.0158
1.0553
0.0538
28.1516
0.5
0.4547
0.0453
9.0560
0.0304
14.9540
−1
−1.0416
0.0416
4.1616
0.0476
−21.8600
−0.8
−0.8728
0.0728
9.1037
0.0445
−19.5951
−1
−1.0115
0.0115
1.1490
0.0158
−63.8326
−0.8
−0.8107
0.0107
1.3424
0.0128
−63.5246
0.5
0.5380
0.0380
7.6064
0.0234
22.9537
0.2
0.1858
0.0142
7.0770
0.0186
10.0106
0.2
0.1877
0.0123
6.1475
0.0187
10.0234
0.5
0.4891
0.0109
2.1756
0.0232
21.0530
0.5
0.5103
0.0103
2.0632
0.0212
24.0984
0.2
0.1997
0.0003
0.1565
0.0204
9.7718
1
—
—
—
—
—
—
0
—
—
—
—
—
—
0
—
—
—
—
—
—
0
—
—
—
—
—
—
1
—
—
—
—
—
—
0.6
0.5956
0.0044
0.7365
0.0526
11.3276
0.6
0.6407
0.0407
6.7772
0.0460
13.9313
1.36
1.4008
0.0408
2.9999
0.1074
13.0425
0.6
0.6939
0.0939
15.6422
0.0779
8.9014
1.36
1.4723
0.1123
8.2542
0.1007
14.6142
Overall mean value
0.0266
4.2691
0.0371
Note: “—” indicates that the corresponding parameter is fixed at the true value.