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 value
Estimate
Abs. bias
APB (%)
Std. err.
t-test
value
0.5
0.4821
0.0179
3.5760
0.0175
27.4890
0.6
0.5734
0.0266
4.4307
0.0210
27.2577
0.5
0.5622
0.0622
12.4302
0.0540
10.4099
0.6
0.6658
0.0658
10.9583
0.0639
10.4166
0.3
0.3044
0.0044
1.4537
0.0103
29.4308
0.3
0.3060
0.0060
2.0160
0.0306
9.9954
−0.4
−0.4101
0.0101
2.5280
0.0409
−10.0344
1
—
—
—
—
—
—
0
—
—
—
—
—
—
0.6
0.5990
0.0010
0.1662
0.0461
12.9991
0
—
—
—
—
—
—
1
—
—
—
—
—
—
0
—
—
—
—
—
—
0.6
0.6477
0.0477
7.9500
0.1004
6.4540
1
—
—
—
—
—
—
0
—
—
—
—
—
—
1
—
—
—
—
—
—
−1
−1.0068
0.0068
0.6816
0.0115
−87.6890
−1
−0.9671
0.0329
3.2903
0.0161
−60.1612
−1
−1.0076
0.0076
0.7647
0.0131
−76.8664
−1
−0.9873
0.0127
1.2682
0.0373
−26.4510
0.3
0.3147
0.0147
4.9110
0.0125
25.2700
0.4
0.3509
0.0491
12.2775
0.0378
9.2838
0.5
0.4985
0.0015
0.2924
0.0210
23.7831
0.6
0.5906
0.0094
1.5625
0.0793
7.4485
1.5
1.5120
0.0120
0.8000
0.0081
186.2896
1.5
1.4793
0.0207
1.3804
0.0187
79.1925
1.5
1.4993
0.0007
0.0456
0.0145
103.4676
1.5
1.5156
0.0156
1.0393
0.0538
28.1629
0.5
0.4546
0.0454
9.0790
0.0304
14.9559
−1
−1.0417
0.0417
4.1748
0.0476
−21.8780
−0.8
−0.8719
0.0719
8.9935
0.0445
−19.5883
−1
−1.0112
0.0112
1.1215
0.0158
−63.9360
−0.8
−0.8106
0.0105
1.3187
0.0127
−63.6256
0.5
0.5379
0.0379
7.5720
0.0234
22.9583
0.2
0.1859
0.0141
7.0455
0.0184
10.0864
0.2
0.1877
0.0123
6.1600
0.0186
10.1024
0.5
0.4882
0.0118
2.3654
0.0232
21.0692
0.5
0.5094
0.0094
1.8768
0.0211
24.1120
0.2
0.1995
0.0005
0.2665
0.0204
9.7651
1
—
—
—
—
—
—
0
—
—
—
—
—
—
0
—
—
—
—
—
—
0
—
—
—
—
—
—
1
—
—
—
—
—
—
0.6
0.5952
0.0048
0.7927
0.0525
11.3353
0.6
0.6377
0.0377
6.2887
0.0457
13.9456
1.36
1.4002
0.0402
2.9534
0.1073
13.0519
0.6
0.6930
0.0930
15.4950
0.0776
8.9351
1.36
1.4628
0.1028
7.5551
0.1001
14.6143
Overall mean value
0.0262
4.2400
0.0370
Note: “—” indicates that the corresponding parameter is fixed at the true value.