Abstract The existence of the “tree” generalisation of the multinomial logit model, its consistency with theories of “rational” human behaviour, as well as the advantages offered in terms of more realistic modelling by this model form, have all been known for some time. Although a few studies have applied tree logit models, however, applications have been restricted by practical and theoretical difficulties in the estimation of these models, in particular with the sequential estimation that has normally been necessary. This paper presents a new estimation procedure, employing the maximum likelihood criterion, for the estimation of tree logit models. The procedure is based on a simple notation for the tree structure and is believed to be original. It offers the possibility of estimating simultaneously all the parameters of general tree logit structures, without restriction on the number of levels but allowing the user to impose constraints of parameter equality where required. Results are presented indicating both the usefulness of the method in deriving improved models, relative to the more familiar sequential estimations, and its practicality and speed of operation on main-frame, minicomputers and microcomputers.


    Access

    Check access

    Check availability in my library

    Order at Subito €


    Export, share and cite



    Title :

    Estimating “tree” logit models


    Contributors:

    Published in:

    Publication date :

    1986-06-03


    Size :

    17 pages




    Type of media :

    Article (Journal)


    Type of material :

    Electronic Resource


    Language :

    English




    Estimating Nested Logit Models with Censored Data

    Newman, Jeffrey P. / Ferguson, Mark E. / Garrow, Laurie A. | Transportation Research Record | 2013


    Estimating Nested Logit Models with Censored Data

    Newman, Jeffrey P | Online Contents | 2013



    Evaluation of Optimization Methods for Estimating Mixed Logit Models

    Bastin, Fabian / Cirillo, Cinzia / Hess, Stephane | Transportation Research Record | 2005