03-31-2020, 08:08 AM
Specification and estimation discrete choice models
Modelling Transport -
Ortuzar Willumsen
The previous chapter provided an overview of discrete choice modelling and an introduction to different
model forms and theoretical frameworks for individual decisions. This chapter is devoted to a discussion
of two key issues: how to fully specify a discrete or disaggregate model (DM) and how to estimate such
a model once properly specified.
The search for a suitable model specification involves selecting the structure of the model (MNL, NL,
ML, etc.), the explanatory variables to consider, the form in which they enter the utility functions (linear,
non-linear) and the identification of the individual’s choice set (alternatives perceived as available). In
broad terms the objectives of a specification search include realism, economy, theoretical consistency
and policy sensitivity. In other words, we search for a realistic model, which does not require too many
data and computer resources, does not produce counter-intuitive results and is appropriate to the decision
context in which it is to be used. Early aggregate models such as those discussed in Chapters 5 and 6
were often critically portrayed as policy insensitive, either because key variables have been completely
left out of the model or because important model components have been specified as insensitive to
certain policies (e.g. consider the problem of inelastic trip generation). Most of the features of model
specification are susceptible to analysis and experimentation (see Leamer 1978) but they are also strongly
dependent on study context and data availability.
To Read more...
Specification and estimation discrete choice models from Modelling Transport -Ortuzar Willumsen-2.pdf Size: 744.27 KB Downloads: 9
register as member and login to download attachment [pdf]
use for Educational Purposes Only
Modelling Transport -
Ortuzar Willumsen
The previous chapter provided an overview of discrete choice modelling and an introduction to different
model forms and theoretical frameworks for individual decisions. This chapter is devoted to a discussion
of two key issues: how to fully specify a discrete or disaggregate model (DM) and how to estimate such
a model once properly specified.
The search for a suitable model specification involves selecting the structure of the model (MNL, NL,
ML, etc.), the explanatory variables to consider, the form in which they enter the utility functions (linear,
non-linear) and the identification of the individual’s choice set (alternatives perceived as available). In
broad terms the objectives of a specification search include realism, economy, theoretical consistency
and policy sensitivity. In other words, we search for a realistic model, which does not require too many
data and computer resources, does not produce counter-intuitive results and is appropriate to the decision
context in which it is to be used. Early aggregate models such as those discussed in Chapters 5 and 6
were often critically portrayed as policy insensitive, either because key variables have been completely
left out of the model or because important model components have been specified as insensitive to
certain policies (e.g. consider the problem of inelastic trip generation). Most of the features of model
specification are susceptible to analysis and experimentation (see Leamer 1978) but they are also strongly
dependent on study context and data availability.
To Read more...
Specification and estimation discrete choice models from Modelling Transport -Ortuzar Willumsen-2.pdf Size: 744.27 KB Downloads: 9
register as member and login to download attachment [pdf]
use for Educational Purposes Only