graphs/graph_enac_l.gif
English only
INTER  >  TRANSP-OR > Workshop on Discrete Choice Models
Logo TRANSP-OR
Content

Workshop on Discrete Choice Models

August 28 - 30, 2008

Ecole Polytechnique Fédérale de Lausanne, room GC B3 30 (click here for a map)

The 2008 workshop will be organized in the same spirit as the previous ones: an informal meeting for the exhange of ideas around discrete choice models, with the objective to trigger new collaborations, or strenghten existing ones. At the end of the workshop, a list of potential collaborations will be identified, with specific objectives.

Registration

Top

There is no registration fee. Everyone interested is invited to attend. Presentations are upon invitation only. All participants, including speakers, must register with the following form.

Last Name: Required field
First Name: Required field
Email: Required field
Affiliation: Required field
Anti-spam:
    

Schedule

Top
Thursday Aug. 28, 13:45 - Friday Aug. 29, 12:25
Presentations (see detail below)
Friday Aug 29, 14:00 - 17:00
Workshop meeting
Saturday August 30, 9:00 - 12:00
Hiking and paragliding
Saturday August 30, 12:00
Genuine swiss fondue in a chalet up in the mountain
Thursday
13:45 13:55 Welcome
13:55 14:30 Börjesson SweVoT 2007 - A new Swedish Value-of-Time Study
14:30 15:05 Newman Pieces of Choices: Alternative Allocations in Network-GEV Models
15:05 15:40 Ibanez Distribution of Hicksian welfare measures for discrete choice models with imperfectly correlated utilities
15:40 16:10 Break
16:10 16:45 Axhausen What does similarity measure ?
16:45 17:20 Frejinger Adaptive route choices in risky networks
17:20 17:55 Schüssler Using GPS data for route choice modelling
17:55 18:30 Hurtubia The Constrained Multinomial Logit: a semi-compensatory choice model
Friday
09:00 09:35 Robin Behavioural modeling of dynamic facial expression recognition
09:35 10:10 Cherchi A mixed logit mode choice model on panel data: accounting for systematic and random heterogeneity and different correlation over time periods
10:10 10:45 de Lapparent Choice of a screening rule and random utility maximization: an integrated behavioral framework
10:45 11:15 Break
11:15 11:50 Eliasson Variations in valuations of noise and intrusion effects. Are stated choice results compatible with hedonic prices?
11:50 12:25 Ramjerdi Variations of VOT

Accomodation

Top

EPFL-rates apply to the following hotels. Prices are indicative. Please contact the hotel for the exact rates.

In Lausanne

Swiss Youth Hotels
Bois-de-Vaux 36, 1007 Lausanne
tél: +41 21 626.02.22
fax: +41 21 626.02.26
adresse email: lausanne@youthhostel.ch
single: 78.00
Hôtel Elite
Avenue Ste-Luce 1
1003 Lausanne
tél: +41 21 320 23 61
fax: + 41 21 320 39 63
single: 117.00; double: 174.00
Hôtel Alagare
Minotel Suisse
Rue du Simplon 14
1006 Lausanne
tél: +41 21 617 92 52
fax: +41 21 617 92 55
single 105.00; double: 150.00
Hôtel Alpha-Palmiers
Fassbind Hotels
Rue du Petit.Chęne 34
1003 Lausanne
tél: +41 21 555 59 99
fax: +41 21 555 59 98
single: 158.00; double: 220.00

EPFL area

Hotel Pré-Fleuri***
Rue du Centre 1, 1025 St-Sulpice.
Email: prefleuri@bluewin.ch
Tél. 021 691 20 21
Fax 021 691 20 20
Price for a single room around CHF 150.-
Motel des Pierrettes** St-Sulpice
10 minutes walk to EPFL
Route cantonale 19, 1025 St-Sulpice
It has no web-site but you can call at +41 21 691 25 25.
It has no restaurant.
Price for a single room, around CHF 110.- (special price for EPFL hosts)
Hostellerie du Débarcadčre
Chemin du Cręt 7, 1025 St-Sulpice,
It belongs to "Relais& Châteaux"
Web-site: www.debarcadere.ch
Price for a single room around CHF 170.- (special price for EPFL hosts)
Novotel Lausanne Bussigny
35, Route de Condémines, 1030 Bussigny
(15 minutes by car, no bus possibilities)
Price for a single room, around CHF 200.- (special price for EPFL hosts)

Transportation

Top

The easiest way to get to EPFL is to take the train from Geneva Airport to Renens. In Renens, take the light-rail (called M1) towards Lausanne. There is a stop at EPFL. The travel time is about 1 hour.

A map of the bus and metro network can be found here and time tables are available at the Lausanne Transport web page. Note that tickets must be bought before departure in machines only accepting coins. The price for a one-way ticket from the center of Lausanne to EPFL is 2.80 Fr. (two zones) and most machines do not give back change.

Check the Swiss Federal Railways website.

To navigate within EPFL, use map.epfl.ch.

Consult also this page

List of presentations (12)

Top

Click on the title to download the slides (if available).

  • What does similarity measure ? by Kay Axhausen
  • SweVoT 2007 - A new Swedish Value-of-Time Study by Maria Börjesson
  • A mixed logit mode choice model on panel data: accounting for systematic and random heterogeneity and different correlation over time periods by Elisabetta Cherchi
  • Choice of a screening rule and random utility maximization: an integrated behavioral framework by Matthieu de Lapparent
  • Variations in valuations of noise and intrusion effects. Are stated choice results compatible with hedonic prices? by Jonas Eliasson
  • Adaptive route choices in risky networks by Emma Frejinger
  • The Constrained Multinomial Logit: a semi-compensatory choice model by Ricardo Hurtubia
  • Distribution of Hicksian welfare measures for discrete choice models with imperfectly correlated utilities by Nicolas Ibanez
  • Pieces of Choices: Alternative Allocations in Network-GEV Models by Jeffrey Newman
  • Variations of VOT by Farideh Ramjerdi
  • Behavioural modeling of dynamic facial expression recognition by Thomas Robin
  • Using GPS data for route choice modelling by Nadine Schuessler

List of abstracts

Top
  1. What does similarity measure ? by Kay Axhausen (IVT, ETH Zürich)

    This presentation will discuss the various ways in which the field has tried to capture systematically and will try to link these measures to a number possible behavioural effects, which do not allow a clear prediction of the sign or size of the parameter of the similarity variable. The paper will review a number of studies of route choice (public transport, cycling, freight) undertaken in Zürich, which incorporated explicit similarity measures.

    Click here to download the presentation

  2. SweVoT 2007 - A new Swedish Value-of-Time Study by Maria Börjesson (WSP Sweden)

    WSP are currently performing a new Swedish Value-of-Time Study implying estimation of the value of time for private trips, by using Stated Choice techniques. The values of time currently in use rely on a survey study from 1994. The new study will up-date the values of time by making use of the methodological developments since 1994 and by in detail investigate how the value of time among drivers has developed between 1994 and 2007. This presentation will focus on the latter issue. In November 2007 a following up of the survey from 1994 (oct/nov) for drivers was carried out. Survey design, sampling- and interview method and even the same field work institute were identical between the survey years. The development of value of time is investigated using different model assumptions and is compared to inflation and income growth. Initial results imply that value of time has increased more than what income growth and inflation explain. Some preliminary results from the pilot surveys of the main study (including all travel modes) will also be discussed in the presentation, for instance differences in the value of time distribution between samples interviewed by telephone and internet.

    Click here to download the presentation

  3. A mixed logit mode choice model on panel data: accounting for systematic and random heterogeneity and different correlation over time periods by Elisabetta Cherchi (University of Cagliari (Italy))

    Travel demand models are usually estimated using cross-sectional data. However, the problem of whether a one-day data set is able to properly describe travel choices has always concerned the researchers. Panel data were extensively used around the 80s mainly to explore sample designs, measurement errors, time-using, dynamic effect over waves. Then they have been re-discovered only recently, with the increased need of understanding variability in travel behavior and the parallel advances in travel demand modeling techniques. The most recent works on panel data are based on multi-week travel diary, where all the daily trips made by all the members of a limited number of households are continuatively recorded over a certain number of weeks. Multi-week data sets have been used to detect rhythms of daily life, to compare different indexes that measure similarities of travel behavior, to draw evidence on the parametric assumptions behind the value of time distribution, to explain current behavior on the individuals’ history and experience, to estimate discrete-continuous models of mode choice and activity duration and to examine time-use in children’s activities. Although several multi-day travel surveys and more advanced modelling techniques are now available, analyses based on panel data are still not a standard tool for travel behaviour modelling and demand forecasting. In particular, models estimated on panel data raise new interesting issues such as the several dimensions of the correlation across responses, the stability/variability of individual behavior over days and over weeks and the possibility to study the effect of repeated observations over days and weeks. Using a six week panel data several mode choice models are estimated accounting for interpersonal variability (systematic and random heterogeneity over individual preferences and responses) and different dimensions of correlation across responses. The stability of taste preferences over days and weeks, the effect of repeated observations and the effect of the correlation across individuals over three time periods (the single day, the single week, the days of a wave), is investigated. We found that taste preferences are quite stable over days and weeks (except for Friday and for the last week of survey). We also found that accounting for correlation over individual improves model results, however the degree of the correlation and its effect on the significance of systematic and heterogeneity differs over the time periods considered. Moreover, the improvement in model estimates achieved by accounting for correlation across individuals is to a great extent given by persistent behaviors.

    Click here to download the presentation

  4. Choice of a screening rule and random utility maximization: an integrated behavioral framework by Matthieu de Lapparent (INRETS)

    As observer of a discrete choice situation a decision maker is faced to, one knows only the set of available choices, their observed characteristics, her ones, and the final choice. Even if information that is provided to the decision maker is perfect, she may use various rules to simplify decision tasks, especially if they are complicated. In this particular instance, one assumes that random utility maximization is the behavior that underlies observed choices but one accepts that one does not know whether all the choice alternatives are taken into account or not for final decision. One may assume that the decision maker may select a non empty subset of alternatives from her available set before proceeding to her final choice. She may screen some of them out. Screening rules restrict actually the set of choice alternatives that are evaluated for final decision. A quick look at the existing literature shows that it has been proposed several approaches to model preselection of choice alternatives, e.g. conjunctive and disjunctive rules, random bounds, univariate and multivariate latent consideration levels, and penalized utility function. The present approach proposes a dynamic behavioral framework that models a sequence of three nested discrete choices, the latter being the only observed: the choice of a screening rule, the choice of a considered set of alternatives, and the choice of a final alternative. All of them are based on latent behavioral processes that mix together to produce final decision. The outline of the presentation if as follows. First, the approach is put into perspective with respect to existing approaches. After a detailed description of the different considered screening rules, it is then proposed an operational model that generalizes RUM theory. It concludes on its advantages and drawbacks.

    Click here to download the presentation

  5. Variations in valuations of noise and intrusion effects. Are stated choice results compatible with hedonic prices? by Jonas Eliasson (Förestĺndare Centrum för transportstudier, KTH)

    Valuation of intrusion effects (noise, visual intrusion etc.) from roads and railways can be estimated either using market data (hedonic price studies) or hypothetic data (contingent valuation, stated choice etc.). In general, market-based methods give lower valuation than hypothetic methods, since market-based methods will measure the valuation of the “marginal buyer” – the market segment demanding the least “rebate” for intrusion effects – while hypothetic methods will (in general) measure average valuations. Self-selection will thus cause market valuations of intrusion effects to be lower than average valuations. This distinction has been pointed out before, but seems to be overlooked in many cases, especially in empirical applications. The main contribution of the present study is to explore the variation of valuations across the population, and try to confirm the theoretical expectation that the valuation of the marginal buyers (those with leat willingness to pay to avoid intrusion effects) should converge to the valuation typically found in hedonic price studies (or other market-based studies). If not, there may be other differences between hypothetic methods and market-based methods, perhaps inherent in the methodologies. We try to answer this question by exploring the preference variation among respondents in a stated choice setting, where recent house buyers choose between two hypothetic alternative residences. Preference variation is explored using choice-based segmentation (based on least accepted distance to a road or railway) and using random-parameters logit models. The two methods yield comparable but not identical results. We show that there are indeed very large differences across the population regarding the willingness to pay to avoid intrusion effects. The results of the marginal market segment (those with least willingness to pay to avoid intrusion) seem indeed to converge towards valuations normally found in hedonic price studies. However, there are some signs that market-based valuations are lower and price decreases due to intrusion effects abate faster. It is possible that this is due to respondents tending to overestimate how far from roads and railways intrusion effects are present, despite the fact that we (for some respondents) used films and soundtracks to illustrate noise and visual intrusion. A second contribution is the comparison of parametric non-linear estimation (Box-Cox) and a method akin to semi-parametric estimation - a weighted average of a family of piecewise linear functions. Both methods give similar results, showing that intrusion effects are highly non-linear, and extend far beyond the immediate vicinity of the road. Thirdly, we compare intrusion effects from three types of infrastructure: motorways, railways and major urban roads. Railway valuations are shown to be similar to motorways, while valuations of proximity to urban roads are considerably lower.

    Click here to download the presentation

  6. Adaptive route choices in risky networks by Emma Frejinger (TRANSP-OR, EPFL)

    In this talk we present the first econometric specification for a routing policy choice model with a non-expected utility function to capture travelers' route adaptation to real-time information and risk attitudes. A routing policy is a decision rule that maps from all possible network states to next links out of decision nodes (dissertation of Song Gao, 2005). Travelers' attitudes toward risk are modeled by utilizing both expected utility (EU) and non-expected utility (non-EU) theory. We estimate routing policy choice and non- adaptive path choice models based on synthetic data and compare the models in terms of prediction results.

    Click here to download the presentation

  7. The Constrained Multinomial Logit: a semi-compensatory choice model by Ricardo Hurtubia (TRANSP-OR EPFL)

    The traditional formulation of logit models assumes a compensatory (indirect) utility function in which consumer’s strategy assumes a trade-off between attributes. This approach fails to recognize attribute thresholds in consumer behavior, or a more generic domain where such a compensatory strategy is contained. A mixed strategy is proposed, which combines the compensatory strategy valid in the interior of the choice domain with cutoff factors that restrain choices to the domain edge. The proposed model combines the multinomial logit model with a binomial logit factor that represents soft cutoffs. This approach extends previous contributions in several ways and allows multiple dimensions for cutoff factors. In addition to considering individual behavior, it introduces system constraints such as capacity and inter-agent interactions (choice externalities).

    Click here to download the presentation

  8. Distribution of Hicksian welfare measures for discrete choice models with imperfectly correlated utilities by Nicolas Ibanez (Institute for Transport Studies, University of Leeds, UK)

    The requirements imposed upon discrete choice models to ensure consistency with random utility maximisation (RUM) have attracted the attention of researchers ever since the pioneering theoretical work on RUM in the 1960s. Noting RUM’s influential position as a behavioural paradigm to explain individual choices, this paper seeks to illuminate deeper understanding of the requirements governing the use of RUM demand models for welfare analysis, that is, for the evaluation of the effects on society of a given policy affecting consumer behaviour. In calculating these effects the aim is to use measures that keep consistency with the demand model being used. The most commonly employed measures (‘rule-of-a-half’ and ‘logsum’) have proved inaccurate when the policy under study has a non negligible impact on the income available to individuals, for instance, when considering in a transport context policies that would change car ownership or that would implement extensive road pricing schemes. Hence, the paper discusses alternative (Hicksian) measures of benefit that might be more appropriate when this income effect is significant. The formulation of welfare measures within a Hicksian scheme has been previously analysed in the literature under three different approaches: those focused on building aggregate measures based on the behaviour of a representative consumer for the population under study (Hau, 1985), those based on the direct simulation of the disaggregate random utilities of each consumer in the population (McFadden, 1999), and those that derive expressions for the moments of the distributions of welfare measures inherent to each consumer in terms of one-dimensional integrals that only depend on choice probabilities (Karlström, 1998, Palma and Kilani, 2003). The three approaches rely on the perfect correlation between the random components of the utilities before and after a policy intervention. This assumption would, intuitively at least, seem rather strong. However, the relaxation of this constraint is not supported by the latter of the approaches and increases the computational burden of the other two. The paper elaborates on the relaxation of this assumption of perfect correlation, and discusses its effect on the distribution of welfare measures for each consumer. In doing so, the analysis allows not only for possible uncertainty across choices from the same consumers, as originally postulated by Manski (1977), but also for differing effects of this uncertainty on alternatives that are affected by the policy and alternatives that are not. In summary, the paper adds to the current literature on the validity of RUM travel demand models for welfare analysis in the presence of income effect, and in particular offers new procedures for relaxing the assumption of correlated errors before and after a policy intervention.

    Click here to download the presentation

  9. Pieces of Choices: Alternative Allocations in Network-GEV Models by Jeffrey Newman (TRANSP-OR, EPFL)

    Generalized extreme value (GEV) models provide a convenient way to model choice behavior that is consistent with utility maximization theory, but the development of specific new models within the GEV family has been slow, due to the difficulty of ensuring new formulations comply with all the GEV rules. The network GEV structure provides a tool to quickly generate new models in the GEV family, without the burden of complex analysis of the new model to ensure its properties. However, such models contain an unwieldy number of parameters, and require numerous restrictions to allow for model identification. This paper will examine an upside-down logit methodology for imposing normalization restrictions on allocations, which can also incorporate disaggregate data into the structure of the correlation among alternatives. We will also explore a “all zero nests” model, where the logsum parameters of every nest are zero, and the levels of correlation are entirely controlled by the allocation parameters.

    Click here to download the presentation

  10. Variations of VOT by Farideh Ramjerdi (Institute of Transport Economics, Norwegian Center for Transport Research)
  11. Behavioural modeling of dynamic facial expression recognition by Thomas Robin (TRANSP-OR EPFL)

    Facial expression recognition is a hard and ambiguous problem in computer vision. It is hard due to the wide variety of faces and the wide variety of ways of representing the same expression. And it is ambiguous because, even though common approaches treat it as a classification problem, actually when looking at the same scene, different people can feel a different expression. We will present a new model framework for facial expression recognition in video sequences, inspired from car lane changing and discrete choice models. It is divided into two parts: the first one which is the dynamic evaluation of each potential expression, updated at each time step (one second seems to be appropriate). The second consists in linking the expression choice with the dynamic evaluation. In order to estimate the model by likelihood maximization, an internet survey is conducted at the moment to obtain expressions labels on facial videos. The sequences are taken from the Cohn-Kanade and the FEED data bases.

    Click here to download the presentation

  12. Using GPS data for route choice modelling by Nadine Schuessler (IVT, ETH Zurich)

    GPS data has gained considerable importance in transportation research over the last decade due to its manifold advantages compared to classic survey methods. In particular route choice modelling profits from the substantially improved levels of detail and accuracy, both spatially and temporally. On the other hand, the new technology implicates new challenges concerning data processing, especially in the case of a large scale person-based GPS study. The presentation will give an overview about the work on such a large-scale person-based GPS study that has been recently undertaken at IVT. In total, the GPS-records of 4878 participants were obtained each of whom carried an on-person GPS receiver for 6.64 days on average. After a short introduction with regard to the data processing challenges, the focus will be laid on the resulting route choice models for bike, car and public transport and the inclusion of similarity factors these models.

    Click here to download the presentation

List of participants (22)

Top
  • Araújo Rogério, Universidade Federal do Ceará
  • Axhausen Kay, IVT, ETH Zürich
  • Ben-Elia Eran, University of Utrecht, Netherlands
  • Bierlaire Michel, TRANSP-OR, EPFL
  • Börjesson Maria, WSP Sweden
  • chaaba fadhila, bh
  • CHEN Jingmin, TRANSP-OR EPFL
  • Cherchi Elisabetta, University of Cagliari (Italy)
  • Cyganski Rita, German Aerospace Center (DLR)
  • de Lapparent Matthieu, INRETS
  • Eliasson Jonas, Förestĺndare Centrum för transportstudier, KTH
  • Flötteröd Gunnar, Berlin Institute of Technology
  • Frejinger Emma, TRANSP-OR, EPFL
  • Hurtubia Ricardo, TRANSP-OR EPFL
  • Ibanez Nicolas, Institute for Transport Studies, University of Leeds, UK
  • Newman Jeffrey, TRANSP-OR, EPFL
  • Ramjerdi Farideh, Institute of Transport Economics
  • Ramjerdi Farideh, Institute of Transport Economics
  • Ramjerdi Farideh, Institute of Transport Economics, Norwegian Center for Transport Research
  • Robin Thomas, TRANSP-OR EPFL
  • Schuessler Nadine, IVT, ETH Zurich
  • Verstraete Adriana, EPFL - ENAC