No open-access location is known for this record yet.
No open-access location is known for this record yet.
Transport planning forecasts have often used members of the logit model family for a long time. Although the parameters of logit models are usually estimated by the method of maximum likelihood, this paper is based on the estimation of logit model parameters from data sets formed by simulation, where the underlying choice processes are under control. It discusses what kind of errors, about individual assessment as well as prediction of policy effects, would result from the use of the estimated model(s). It also considers how the answer to this question depends on basic model imperfections and factors such as sample size. The estimated models and the underlying data set, considered here, concern the modal split for regional commuting, but the results obtained about modelling errors and problems seem to be much more general and to apply to all types of probit model use. The simulation method used had four steps. Estimates of parameters and the variance of these estimates are then discussed. Illustrative examples considered include the distribution of estimated effects from a suggested policy, and to show the influence of the level of randomness. Finally, some effects of model imperfections are discussed, including the use of the quasi-T-test for overspecification, and taste variation. For the covering abstract, see IRRD E101013.
Loading…
Imported from OpenAlex (CC0) — nothing here is generated.