03-31-2020, 07:59 AM
Discrete choice models
Modelling Transport
-Ortuzar Willumsen
In this chapter we provide a comprehensive introduction to discrete choice (i.e. when individuals have to
select an option from a finite set of alternatives) modelling methods.We start with some general considerations
and move on to explain the theoretical framework, random utility theory, in which these models are
cast. This serves us to introduce some basic terminology and to present the individual-modeller ‘duality’
which is so useful to understanding what the theory postulates. Next we introduce the two most popular
discrete choice models: Multinomial and Nested Logit, which taken as a family provides the practitioner
with a very powerful modelling tool set. We also discuss other choice models, in particular Mixed Logit
which is now recognised as the standard in the field, and also consider the benefits and special problems
involved when modelling with panel data and when one wants to incorporate latent variables. These are
two increasingly important subjects and should shortly become standard practice. Finally, we briefly look
at other paradigms which offer an alternative perspective to the classical utility-maximising approach.
The problems of model specification and estimation, both with revealed- and stated-preference data,
are considered in sufficient detail for practical analysis in Chapter 8; we provide information about
certain issues, such as validation samples, which are seldom found in texts on this subject. The problem
of aggregation, from various perspectives, and the important question of model updating and transference
(particularly for those interested in a continuous planning approach to transport), are tackled in Chapter 9.
To Read more...
register as member and login to download attachment [pdf]
use for Educational Purposes Only
Modelling Transport
-Ortuzar Willumsen
In this chapter we provide a comprehensive introduction to discrete choice (i.e. when individuals have to
select an option from a finite set of alternatives) modelling methods.We start with some general considerations
and move on to explain the theoretical framework, random utility theory, in which these models are
cast. This serves us to introduce some basic terminology and to present the individual-modeller ‘duality’
which is so useful to understanding what the theory postulates. Next we introduce the two most popular
discrete choice models: Multinomial and Nested Logit, which taken as a family provides the practitioner
with a very powerful modelling tool set. We also discuss other choice models, in particular Mixed Logit
which is now recognised as the standard in the field, and also consider the benefits and special problems
involved when modelling with panel data and when one wants to incorporate latent variables. These are
two increasingly important subjects and should shortly become standard practice. Finally, we briefly look
at other paradigms which offer an alternative perspective to the classical utility-maximising approach.
The problems of model specification and estimation, both with revealed- and stated-preference data,
are considered in sufficient detail for practical analysis in Chapter 8; we provide information about
certain issues, such as validation samples, which are seldom found in texts on this subject. The problem
of aggregation, from various perspectives, and the important question of model updating and transference
(particularly for those interested in a continuous planning approach to transport), are tackled in Chapter 9.
To Read more...
register as member and login to download attachment [pdf]
use for Educational Purposes Only