Discrete choice modelling and stated choice survey design (CIV5127Z)
Course convenor: A/Prof Mark Zuidgeest (Department of Civil Engineering)
Motivation for the course
Discrete choice modelling offers a statistical technique typically used in marketing research, which has more recently drawn the attention of transport specialists. Choice models model the decision process of an individual or a market segment in a particular context. It looks at choices that customers make between products or services (a new BRT transport service for example) and use these to identify systematic patterns in observed or hypothetical choices. The estimated models can for example be used to predict how consumers respond to competing products and estimate the impact of product innovation or changes in income on the share of public transport. Stated choice surveys provide a survey methodology for investigating hypothetical travel behaviour. Stated choice survey design looks at how to collect the data for efficient discrete choice model estimation with as little bias as possible.
South African issues related to innovation and modernization in the transportation, health and other sectors demand knowledge on choice behaviour of consumers and trip makers alike and better understanding of the ways to model choice behaviour and collect data to estimate these models. The course therefore covers all the steps required for successful choice modelling analyses, from inception via survey design and data collection to modelling and implementation.
The objectives of this course are: to provide participants with an introduction to the theory of discrete choice models and their data requirements; to train participants in the setting-up and estimation of various types of discrete choice models using commonly available software; to learn to use and appraise the results of a discrete choice model and how to use them for forecasting; to present various case studies of discrete choice methods in the African context (transport, health economics etc.); to provide participants with an introduction to the theory of stated choice surveys; to train participants in the (efficient) design of stated choice surveys; to learn to use and appreciate the development of efficient design in a model like Ngene; and to present various case studies of stated choice survey design in the African context (transport, health economics etc.).
Course structure and content
The course comprises four phases:
- a pre-contact period of five weeks, involving some 30-40 hours of preparatory reading and assignments;
- a week of intensive contact time at UCT, comprising 40-50 hours of formal lectures and class assignments;
- a two-hour course test (on the Monday following the contact week), intended to evaluate students’ understanding of selected aspects of the material they have been exposed to during the contact week, and
- a post-contact period of seven weeks, involving an assignment or assignments requiring about 100 hours of work.
The material presented during the contact period is structured around the following broadly sequential themes or topics:
- Discrete choice modelling: Multinomial Logit Model; Model estimation; Interpreting model results, appraisal & forecasting; Specification testing; Nested Logit; other GEV models; Mixed Logit, Latest Class models; Random Utility models. Exercises on Multinomial Logit, Mixed Logit, GEVE estimation and model fitting; Case studies in Africa
- Stated choice survey design: Stated Choice Surveys, Fractional factorial design; Orthogonal design; Survey examples; Issues with orthogonal design; efficient design. Exercises on generating efficient designs using Ngene. Case studies in Africa.
Formal presentations on these topics by both Programme staff and invited external specialists are interspersed with work on a group assignment, the product of which is presented and discussed on the last day of the course.
Readings for the preparatory assignment are issued about five weeks prior to the commencement of the contact period. Additional readings may be issued during the contact week or subsequently through the Vula on-line course worksite, together with the brief for the major post-contact phase assignment.
Contact week attendance
Students are expected to attend the contact week on a full-time basis, which will require them to be resident in Cape Town for its duration given that the daily timetable during this period will generally occupy the hours from 08h00 to 17h00. Attendance for the entire contact period is a requirement, and students who are absent for more than 20% of the contact period duration will not be allowed to complete the course.
Students who have completed the course successfully should:
• understand the principles of discrete choice modelling;
• understand the principles of stated choice surveys;
• know emerging data collection methods in the field;
• be able to set-up and measure a discrete choice model;
• be able to develop an efficient stated choice experiment; and
• appreciate the application of discrete choice modelling techniques and that of stated choice survey design in in various sectors and in the African context.
The aggregate mark for this course is compiled as follows:
|minor assignment based on preparatory reading||15%|
|group contact week assignment||10%|
|major post-contact phase assignment||50%|
Students who do not obtain an aggregate mark of at least 50% for their preparatory assignment, course test and post-contact phase assignment will be deemed to have failed the course.