Mixing distributions of tastes with a Combination of Nested Logit (CoNL) kernel: Formulation and performance analysis. (November 2020)

• Record Type: Journal Article
• Authors: Tinessa, Fiore; Marzano, Vittorio; Papola, Andrea
• Is Part Of: Transportation research. Volume 141(2020)
• Page Start: 1
• Page End: 23
• Publication Date: 2020-11
• Subjects: CoNL -- Mixing distribution for tastes -- Latent class -- Random coefficients; Transportation -- Research -- Periodicals; Transportation -- Mathematical models -- Periodicals
• DOI: 10.1016/j.trb.2020.08.007
• Languages: English
• ISSNs: 0191-2615

Abstract

Highlights: Testing the Combination of Nested Logit (CoNL) with stochastic tastes distributions. Various model formulations illustrated with their mathematical properties. Several alternative models identified for comparison. Cross-validation exercise on a real SP mode choice dataset: Naples-Milan corridor. Proposed models outperform existing ones in terms of in-sample/out-of-sample GOF. Abstract: This paper explores the potential of a special instance of the Combination of Random Utility Models (CoRUM; Papola, 2016), termed Combination of Nested Logit (CoNL), as kernel model in conjunction with several types of mixing distributions of tastes (parametric, nonparametric, semiparametric). Various model formulations are illustrated with their mathematical properties, and several alternative kernel models are identified for comparison. An estimation exercise is presented on a real mode choice dataset from a stated preference survey on the intercity corridor between Naples and Milan in Italy. Results, in terms of both in-sample and out-of-sample goodness-of-fit on a 10-fold cross-validation show that models with the proposed CoNL kernel outperform contrasted models.