Montevideo journey-to-work flows, 2016: A doubly constrained gravity model with random effects
Montevideo journey-to-work flows, 2016: A doubly constrained gravity model with random effects

Summary/Abstract: In this study, a doubly constrained gravity model was estimated using generalised linear mixed-effects models for journey-to-workflows in Montevideo, Uruguay. Under the mixed-models framework, Poisson and negative binomial regression models were estimated, finding a better fit for the last distribution. The negative binomial distribution used in the model specification improves the parameter estimation by up to15%. The results are compared with a generalised linear model (GLM) specification, showing that considering regions as fixed effects are insufficient to account for dependence among flows. Modelling spatial autocorrelation by including random effects enables compliance with the no-correlation assumption between residuals while still considering different scenarios for spatial autocorrelation in flows: at the origin,destination, and both. Considering regions of origin and destination as random effects could be a solution for the spatial autocorrelation inflows, and in some cases more complex modelling strategies can be avoided.

• Details
• Contents