The performances of different simulation-based estimation techniques for mixed logit modeling are evaluated. A quasi–Monte Carlo method (modified Latin hypercube sampling) is compared with a Monte Carlo algorithm with dynamic accuracy. The classic Broyden–Fletcher–Goldfarb–Shanno (BFGS) optimization algorithm line-search approach and trust region methods, which have proved to be extremely powerful in nonlinear programming, are also compared. Numerical tests are performed on two real data sets: stated preference data for parking type collected in the United Kingdom, and revealed preference data for mode choice collected as part of a German travel diary survey. Several criteria are used to evaluate the approximation quality of the log likelihood function and the accuracy of the results and the associated estimation runtime. Results suggest that the trust region approach outperforms the BFGS approach and that Monte Carlo methods remain competitive with quasi–Monte Carlo methods in high-dimensional problems, especially when an adaptive optimization algorithm is used.
Get full access to this article
View all access options for this article.
References
1.
McFaddenD., and TrainK.. Mixed MNL Models for Discrete Response.Journal of Applied Econometrics, Vol. 15, 2000, pp. 447–470.
2.
TrainK.Discrete Choice Methods With Simulation.Cambridge University PressCambridge, United Kingdom,2003.
3.
HessS., TrainK., and PolakJ. W.. On the Use of a Modified Latin Hyper-cube Sampling (MLHS) Approach in the Estimation of a Mixed Logit Model for Vehicle Choice.Transportation Research B (forthcoming).
4.
BastinF., CirilloC., and TointP. L.. A Trust Region Monte Carlo Algorithm for Mixed Logit Estimation.Transportation Research GroupFacultés Universitaires Notre-Dame de la Paix, Namur, Belgium,2004.
5.
WalkerJ. L.Extended Discrete Choice Models: Integrated Framework, Flexible Error Structures, and Latent Variables. Ph.D. thesis. Massachusetts Institute of Technology, Cambridge, 2001.
6.
ReveltD., and TrainK.. Mixed Logit with Repeated Choices.Review of Economics and Statistics, Vol. 80, 1998, pp. 647–657.
7.
LeeL.Asymptotic Bias in Simulated Maximum Likelihood Estimation of Discrete Choice Models.Econometric Theory, Vol. 11, 1995, pp. 437–483.
8.
BhatC. R.Quasi-Random Maximum Simulated Likelihood Estimation of the Mixed Multinomial Logit Model.Transportation Research B, Vol. 35, No. 7, 2001, pp. 677–693.
BhatC. R.Simulation Estimation of Mixed Discrete Choice Models Using Randomized and Scrambled Halton Sequences.Transportation Research B, Vol. 37, No. 9, 2003, pp. 837–855.
11.
HessS., PolakJ. W., and DalyA. J.. On the Performance of Shuffled Halton Sequences in the Estimation of Discrete Choice Models. Presented at European Transport Conference 2003, Strasbourg, France, October 8–10, 2003.
12.
GarridoR. A.Estimation Performance of Low Discrepancy Sequences in Stated Preferences. Presented at 10th International Conference on Travel Behaviour Research, Lucerne, Switzerland, August 10–14, 2003.
13.
SándorZ., and TrainK.. Quasi-Random Simulation of Discrete Choice Model.Transportation Research B, Vol. 38, No. 4, 2004, pp. 313–327.
14.
L'EcuyerP., and LemieuxC.. Recent Advances in Randomized Quasi-Monte Carlo Methods. In International Series in Operations Research and Management Science, Vol. 46: Modeling Uncertainty.Kluwer Academic, Boston, Mass., 2002, pp. 419–474.
15.
BastinF., CirilloC., and TointP. L.. Convergence Theory for Non-convex Stochastic Programming With an Application to Mixed Logit.Mathematical Programming B (forthcoming).
16.
ConnA. R., GouldN. M., and TointP. L.. Trust-Region Methods.SIAM Publications, Philadelphia, Pa., 2000.
17.
BastinF., CirilloC., and TointP. L.. An Adaptive Monte Carlo for Computing Mixed Logit Estimators.Journal of Computational Management (forthcoming).
18.
PolakJ. W., AxhausenK. W., and ErringtonT.. The Application of CLAMP to the Analysis of Parking Policy in Birmingham City Centre. Working Paper 554. Transport Studies Group, Oxford University, Oxford, United Kingdom,1990.
19.
PolakJ. W., and AxhausenK. W.. The Birmingham CLAMP Stated Preference Survey. Second Interim Report to Birmingham City Council. Transport Studies Unit, Oxford University, Oxford, United Kingdom, 1989.
20.
HessS., and PolakJ. W.. Mixed Logit Estimation of Parking Type Choice. Presented at 83rd Annual Meeting of the Transportation Research Board, Washington, D.C., 2004.
21.
AxhausenK. W., ZimmermannA., SchönfelderS., RindsfüserG., and HauptT.. Observing the Rhythms of Daily Life: A Six-Week Travel Diary.Transportation, Vol. 29, No. 2, 2002, pp. 95–124.
22.
BhatC. R., and SinghS. K.. A Comprehensive Daily Activity-Travel Generation Model System for Workers.Transportation Research A, Vol. 34, No. 1, 2000, pp. 1–22.
23.
CirilloC., and TointP. L.. A Multinational Comparison of Travel Behaviour Using an Activity-Based Approach.Transportation Research GroupFacultés Universitaires Notre-Dame de la Paix, Namur, Belgium,2001.
NocedalJ., and WrightS. J.. Numerical Optimization.Springer, New York, 1999.
26.
ZhuC., ByrdR. H., and NocedalJ.. Algorithm 778: L-BFGS-B, FORTRAN Routines for Large Scale Bound Constrained Optimization.Transactions on Mathematical Software, Vol. 23, No. 4, 1997, pp. 550–560.
27.
MoreJ. J., and ThuenteD. J.. Line Search Algorithms with Guaranteed Sufficient Decrease.Transactions on Mathematical Software, Vol. 20, No. 3, 1994, pp. 286–307.
28.
VicenteL. N.A Comparison Between Line Searches and Trust Regions for Nonlinear Optimization.Investigação Operacional, Vol. 16, 1996, pp. 173–179.
29.
CirilloC., and AxhausenK. W.. Arbeitsbericht Verkehrs- und Raumplanung, No. 212: Evidence on the Distribution of Values of Travel Time Savings From a Six-Week Diary.Institut für Verkehrsplanung und Transporttechnik, Strassen-und Eisenbahnbau, Eidgenössische Technische Hochschule (ETH) Zürich, Zurich, Switzerland, 2004.
30.
HessS., BierlaireM., and PolakJ. W.. Mixed Logit Estimation of Value of Travel-Time Savings.Transportation Research A, Vol. 39, No. 2–3, 2005, pp. 221–236.
