Using semi-open questions to integrate perceptions in choice models
Glerum, Atasoy, Bierlaire, 2014, in Journal of Choice Modelling
doi:10.1016/j.jocm.2013.12.001
Location |
Low density Swiss areas |
Population |
General |
Sample size |
1763 |
Factor analysis type |
none, none rotation |
Stepwise regression |
yes |
Removal of insignificant variables |
yes |
Reviewed by |
LCM |
Abstract
This research investigates the measurement of perceptions by means of adjectives freely reported by respondents in semi-open questions. It involved the use of semi-open responses of 1763 Swiss individuals to develop indicators for a latent variable representing the perception of comfort of public transportation. The indicators are then incorporated into a discrete choice model of revealed mode choices. Perceptions are assumed to impact choice significantly and this research aims at capturing their complexity using adjectives and integrating them into the hybrid choice modeling framework. We exploit a quantification of the adjectives performed by external evaluators. Given the subjectivity that is involved, we analyze the sensitivity of the results across evaluators who rated the adjectives. We observe that the aggregate indicators of demand, such as market shares, elasticities and values of time, are rather robust across evaluators. This is not the case for the disaggregate indicators that may vary substantially across evaluators. © 2013 Elsevier Ltd.
Factors
Models
Dependent variable |
Mode choice |
Model type |
Logit |
Sample size |
2265.0 |
R2 |
0.443 |
Adjusted R2 |
|
Pseudo R2
(nan)
|
nan |
AIC |
nan |
BIC |
nan |
Log-likelihood at zero |
nan |
Log-likelihood at constants |
nan |
Log-likelihood at convergence |
-1153.0 |
PMM |
Variable |
Coefficient |
p-value |
Constant |
0.423
|
0.022 |
Time |
-0.033
|
0.0 |
Work |
-0.613
|
0.006 |
French |
0.99
|
0.0 |
PT |
Variable |
Coefficient |
p-value |
Constant |
-0.178
|
0.379 |
Time |
-0.06
|
0.001 |
Work |
0.0987
|
0.675 |
French |
-0.228
|
0.542 |
Variable |
Coefficient |
p-value |
Cost |
-0.0658
|
0.0 |
Distance |
-0.236
|
0.0 |
French |
1.11
|
0.66 |
Age < 50 |
1.42
|
0.211 |
Full or part time job |
-8.34
|
0.0 |
Ownersip of 2+ cars |
-7.81
|
0.0 |
Dependent variable |
Mode choice |
Model type |
HCM |
Sample size |
2265.0 |
R2 |
0.425 |
Adjusted R2 |
|
Pseudo R2
(nan)
|
nan |
AIC |
nan |
BIC |
nan |
Log-likelihood at zero |
nan |
Log-likelihood at constants |
nan |
Log-likelihood at convergence |
-1192.0 |
PMM |
Variable |
Coefficient |
p-value |
Constant |
0.416
|
0.024 |
Time |
-0.0313
|
0.0 |
Work |
-0.572
|
0.01 |
French |
0.966
|
0.0 |
PT |
Variable |
Coefficient |
p-value |
Constant |
-0.155
|
0.441 |
Time |
-0.0143
|
0.0 |
Work |
-0.0602
|
0.795 |
French |
-0.073
|
0.81 |
Variable |
Coefficient |
p-value |
Cost |
-0.0637
|
0.0 |
Distance |
-0.233
|
0.0 |
Comfort (PCPT) |
1.06
|
0.001 |
Mean (PCPT) |
3.33
|
0.0 |
French |
-0.559
|
0.072 |
Age < 50 |
-1.3
|
0.0 |
Full or part time job |
-1.1
|
0.0 |
Ownersip of 2+ cars |
-0.73
|
0.002 |
a2 (PCPT indicator) |
-0.243
|
0.008 |
a3 (PCPT indicator) |
-0.473
|
0.0 |
a4 (PCPT indicator) |
-1.14
|
0.0 |
a5 (PCPT indicator) |
-1.23
|
0.0 |
a6 (PCPT indicator) |
-1.22
|
0.0 |
a7 (PCPT indicator) |
-0.391
|
0.0 |
a8 (PCPT indicator) |
-0.56
|
0.0 |
a9 (PCPT indicator) |
-0.842
|
0.0 |
y1 (PCPT indicator) |
0.255
|
0.0 |
y2 (PCPT indicator) |
0.158
|
0.0 |
y3 (PCPT indicator) |
0.108
|
0.026 |
y4 (PCPT indicator) |
0.422
|
0.0 |
y5 (PCPT indicator) |
0.195
|
0.0 |
y6 (PCPT indicator) |
0.218
|
0.0 |
y7 (PCPT indicator) |
0.344
|
0.0 |
y8 (PCPT indicator) |
0.264
|
0.0 |
y9 (PCPT indicator) |
0.237
|
0.0 |
o1 (PCPT indicator) |
-0.0585
|
0.242 |
o2 (PCPT indicator) |
0.255
|
0.0 |
o3 (PCPT indicator) |
0.365
|
0.0 |
o4 (PCPT indicator) |
-0.344
|
0.005 |
o5 (PCPT indicator) |
0.0509
|
0.358 |
o6 (PCPT indicator) |
0.00644
|
0.928 |
o7 (PCPT indicator) |
-0.25
|
0.004 |
o8 (PCPT indicator) |
0.122
|
0.047 |
o9 (PCPT indicator) |
0.18
|
0.019 |
Dependent variable |
Mode choice |
Model type |
HCM |
Sample size |
2265.0 |
R2 |
0.422 |
Adjusted R2 |
|
Pseudo R2
(nan)
|
nan |
AIC |
nan |
BIC |
nan |
Log-likelihood at zero |
nan |
Log-likelihood at constants |
nan |
Log-likelihood at convergence |
-1199.0 |
PMM |
Variable |
Coefficient |
p-value |
Constant |
0.41
|
0.027 |
Time |
-0.0312
|
0.0 |
Work |
-0.56
|
0.011 |
French |
0.969
|
0.0 |
PT |
Variable |
Coefficient |
p-value |
Constant |
-0.132
|
0.509 |
Time |
-0.0293
|
0.0 |
Work |
-0.0928
|
0.689 |
French |
-0.113
|
0.711 |
Variable |
Coefficient |
p-value |
Cost |
-0.0628
|
0.0 |
Distance |
-0.233
|
0.0 |
Comfort (PCPT) |
1.09
|
0.008 |
Mean (PCPT) |
15.7
|
0.0 |
French |
-0.139
|
0.631 |
Age < 50 |
0.0643
|
0.764 |
Full or part time job |
-0.582
|
0.007 |
Ownership of 2+ cars |
-0.362
|
0.114 |
a2 (PCPT indicator) |
0.505
|
0.108 |
a3 (PCPT indicator) |
1.46
|
0.004 |
a4 (PCPT indicator) |
-3.33
|
0.0 |
a5 (PCPT indicator) |
-0.344
|
0.368 |
a6 (PCPT indicator) |
0.742
|
0.093 |
a7 (PCPT indicator) |
-2.3
|
0.0 |
a8 (PCPT indicator) |
1.37
|
0.0 |
a9 (PCPT indicator) |
0.731
|
0.013 |
y1 (PCPT indicator) |
0.116
|
0.0 |
y2 (PCPT indicator) |
0.0742
|
0.0 |
y3 (PCPT indicator) |
0.00423
|
0.897 |
y4 (PCPT indicator) |
0.319
|
0.0 |
y5 (PCPT indicator) |
0.123
|
0.0 |
y6 (PCPT indicator) |
0.051
|
0.077 |
y7 (PCPT indicator) |
0.258
|
0.0 |
y8 (PCPT indicator) |
0.0268
|
0.091 |
y9 (PCPT indicator) |
0.0613
|
0.001 |
o1 (PCPT indicator) |
-1.06
|
0.0 |
o2 (PCPT indicator) |
-0.521
|
0.0 |
o3 (PCPT indicator) |
-0.365
|
0.0 |
o4 (PCPT indicator) |
-1.68
|
0.0 |
o5 (PCPT indicator) |
-0.461
|
0.0 |
o6 (PCPT indicator) |
-0.394
|
0.0 |
o7 (PCPT indicator) |
-1.23
|
0.0 |
o8 (PCPT indicator) |
-0.673
|
0.0 |
o9 (PCPT indicator) |
-0.675
|
0.0 |
Dependent variable |
Mode choice |
Model type |
HCM |
Sample size |
2265.0 |
R2 |
0.427 |
Adjusted R2 |
|
Pseudo R2
(nan)
|
nan |
AIC |
nan |
BIC |
nan |
Log-likelihood at zero |
nan |
Log-likelihood at constants |
nan |
Log-likelihood at convergence |
-1190.0 |
PMM |
Variable |
Coefficient |
p-value |
Constant |
0.419
|
0.023 |
Time |
-0.0323
|
0.0 |
Work |
-0.575
|
0.009 |
French |
0.967
|
0.0 |
PT |
Variable |
Coefficient |
p-value |
Constant |
-0.155
|
0.441 |
Time |
-0.0208
|
0.0 |
Work |
-0.0474
|
0.841 |
French |
-0.0808
|
0.795 |
Variable |
Coefficient |
p-value |
Cost |
-0.0653
|
0.0 |
Distance |
-0.235
|
0.0 |
Comfort (PCPT) |
1.33
|
0.0 |
Mean (PCPT) |
7.47
|
0.0 |
French |
-0.456
|
0.114 |
Age < 50 |
-1.04
|
0.0 |
Full or part time job |
-1.12
|
0.0 |
Ownersip of 2+ cars |
-0.688
|
0.002 |
a2 (PCPT indicator) |
0.00209
|
0.992 |
a3 (PCPT indicator) |
-0.686
|
0.048 |
a4 (PCPT indicator) |
-2.34
|
0.0 |
a5 (PCPT indicator) |
-2.31
|
0.0 |
a6 (PCPT indicator) |
-2.98
|
0.0 |
a7 (PCPT indicator) |
-0.828
|
0.002 |
a8 (PCPT indicator) |
-1.35
|
0.0 |
a9 (PCPT indicator) |
-2.56
|
0.0 |
y1 (PCPT indicator) |
0.23
|
0.0 |
y2 (PCPT indicator) |
0.151
|
0.0 |
y3 (PCPT indicator) |
0.203
|
0.0 |
y4 (PCPT indicator) |
0.433
|
0.0 |
y5 (PCPT indicator) |
0.321
|
0.0 |
y6 (PCPT indicator) |
0.453
|
0.0 |
y7 (PCPT indicator) |
0.324
|
0.0 |
y8 (PCPT indicator) |
0.336
|
0.0 |
y9 (PCPT indicator) |
0.448
|
0.0 |
o1 (PCPT indicator) |
0.0627
|
0.144 |
o2 (PCPT indicator) |
0.388
|
0.0 |
o3 (PCPT indicator) |
0.381
|
0.0 |
o4 (PCPT indicator) |
0.00596
|
0.944 |
o5 (PCPT indicator) |
0.155
|
0.026 |
o6 (PCPT indicator) |
-0.279
|
0.082 |
o7 (PCPT indicator) |
0.0178
|
0.787 |
o8 (PCPT indicator) |
0.27
|
0.0 |
o9 (PCPT indicator) |
0.0496
|
0.631 |