biogeme 2.6a [Mon Apr 17 15:32:48 CEST 2017]

Michel Bierlaire, EPFL

This file has automatically been generated.

Tue Apr 18 19:04:48 2017

Tip: click on the columns headers to sort a table [Credits]

Example of a logit model for a transportation mode choice with 3 alternatives:
- Train
- Car
- Swissmetro, an hypothetical high-speed train
The time coefficient is assumed to be distributed. It is a discrete distribution with two mass points, one at 0, and one at B_TIME_OTHER. The probabilities associated with each mass point are W_0 and W_OTHER, respectively.
Note that the model is unidentifiable. The objective of this example is to illustrate the Biogeme syntax only.
Model: Logit
Number of estimated parameters: 6
Number of observations: 6768
Number of individuals: 6768
Null log likelihood: -6964.663
Init log likelihood: -6964.663
Final log likelihood: -5208.498
Likelihood ratio test: 3512.330
Rho-square: 0.252
Adjusted rho-square: 0.251
Final gradient norm: +9.571e+03
Diagnostic: Normal termination. Obj: 6.05545e-06 Const: 6.05545e-06
Iterations: 20
Run time: 00:00
Variance-covariance: from analytical hessian
Sample file: ../swissmetro.dat

Utility parameters

Name Value Std err t-testp-value
ASC_CAR0.1250.02524.950.00
ASC_SM0.00fixed
ASC_TRAIN-0.3980.0259-15.380.00
B_COST-1.260.0387-32.650.00
B_TIME_00.00fixed
B_TIME_OTHER-2.801.27e+07-0.001.00*
W_00.2511.07e+080.001.00*
W_OTHER0.7491.07e+080.001.00*

Utility functions

IdNameAvailabilitySpecification
1A1_TRAINTRAIN_AV_SPASC_TRAIN * one + B_TIME * TRAIN_TT_SCALED + B_COST * TRAIN_COST_SCALED
2A2_SMSM_AVASC_SM * one + B_TIME * SM_TT_SCALED + B_COST * SM_COST_SCALED
3A3_CarCAR_AV_SPASC_CAR * one + B_TIME * CAR_TT_SCALED + B_COST * CAR_CO_SCALED

Correlation of coefficients

Coefficient1 Coefficient2CovarianceCorrelationt-testp-valueRob. cov.Rob. corr.Rob. t-testp-value
ASC_CAR ASC_TRAIN0.0002570.39518.610.00
ASC_CAR B_COST0.0003280.33736.150.00
ASC_CAR B_TIME_OTHER3.13e-209.79e-260.001.00 *
ASC_CAR W_0-8.17e-22-3.05e-28-0.001.00 *
ASC_CAR W_OTHER-8.48e-22-3.16e-28-0.001.00 *
ASC_TRAIN B_COST0.0002190.21920.840.00
ASC_TRAIN B_TIME_OTHER0.0001083.29e-100.001.00 *
ASC_TRAIN W_0-0.000442-1.60e-10-0.001.00 *
ASC_TRAIN W_OTHER0.0004421.60e-10-0.001.00 *
B_COST B_TIME_OTHER-0.00166-3.37e-090.001.00 *
B_COST W_00.006791.65e-09-0.001.00 *
B_COST W_OTHER-0.00679-1.65e-09-0.001.00 *
B_TIME_OTHER W_0-6.60e+14-0.488-0.001.00 *
B_TIME_OTHER W_OTHER6.60e+140.488-0.001.00 *
W_0 W_OTHER-1.14e+16-1.00-0.001.00 *

User defined linear constraints

1*W_0 + 1*W_OTHER = 1 [1 = 1]

Smallest singular value of the hessian: 1.87584e-17

Unidentifiable model

The log likelihood is (almost) flat along the following combinations of parameters

Sing. value=1.87584e-17
-0.0269223*B_TIME_OTHER
0.463012*W_0
-0.463012*W_OTHER
-0.00501595*Param[9]
0.46228*Param[10]
-0.267461*Param[11]
0.00376281*Param[15]
-0.267461*Param[16]
0.46228*Param[17]