New software for the design of robust stated choice experiments

News - 15 May 2019 - Communication

TU Delft researcher Sander van Cranenburgh, in collaboration with researchers from the University of Sydney, has developed an upgrade for leading software package Ngene. Ngene is used by researchers and professionals to design stated choice experiments. Such experiments are used to gain insight into how people weigh up various aspects when making a choice, for example travel time and travel costs. This software upgrade makes it possible to design statistically accurate choice experiments for so-called regret models, a choice behaviour model developed at Delft.

Choice experiments and behaviour models

‘Choice experiments are widely used in all manner of disciplines to help us to better understand and predict behaviour’, says Sander van Cranenburgh. ‘In a stated choice experiment the respondent is presented with various fictitious choice situations and asked to state which is the most attractive alternative.’ For example: ‘Would you rather take the intercity train to your destination and get there in 90 minutes, or take the high speed train which only takes an hour but costs an extra 10 euros?’. Researchers use special software to create their choice experiments.

Utility model

Researchers setting up choice experiments base these on a behaviour model that describes how people make choices. Until recently that model was always the utility model, a model that assumes that people choose the alternative that will give them the maximum utility (benefit or value). However, this utility model is not always realistic. An alternative behaviour model, developed by Professor Caspar Chorus at TU Delft, assumes that people make choices determined on their urge to minimise regret after the event as far as possible. This regret model regularly proves to be more reliable than the utility model, although this varies according to the context.

Robust

In his earlier research, Sander van Cranenburgh showed that choice experiments created for utility models often do not work well with regret models, and vice versa. That is why he advocates creating robust experiments that work well for more than one behaviour model, such as the regret model and the utility model. To make it easier for researchers to create robust choice experiments, Van Cranenburgh has developed an upgrade for the well-known Ngene software package. The latest version of Ngene, containing this upgrade, will be available this month. An article by Van Cranenburgh on this subject is published in the Journal of Choice Modelling.

More information

Sander van Cranenburg, S.vanCranenburgh@tudelft.nl, +31(0)15-2786957
Inge Snijder, Science Communication Advisor TU Delft, I.Snijder@tudelft.nl, +31 (0)15-2787538

Publication: New software tools for creating stated choice experimental designs efficient for regret minimisation and utility maximisation decision rules - Journal of Choice Modelling
DOI: https://doi.org/10.1016/j.jocm.2019.04.002