Longitudinal investigation of skeletal activity episode timing decisions – A copula approach. (September 2021)

• Record Type: Journal Article
• Authors: Ozonder, Gozde; Miller, Eric J.
• Is Part Of: Journal of choice modelling. Volume 40(2021)
• Publication Date: 2021-09
• Subjects: Skeletal activities -- Start time -- Duration -- Copula -- Longitudinal analysis; Decision making -- Periodicals; Social choice -- Periodicals; Decision making; Social choice; Periodicals; 302.13
• DOI: 10.1016/j.jocm.2021.100306
• Languages: English
• ISSNs: 1755-5345

Abstract

Abstract: This study proposes a joint modelling approach to simulate episode timing decisions of skeletal activities -- activities that typically take precedence in the scheduling process -- using copulas. The analyses focus on examining the correlation between activity start time and duration while exploring the timing behaviour over a 20-year period between 1996 and 2016 in the Greater Toronto and Hamilton Area. For three types of agent classes ("Worker", "Student", "Both"), timing decisions associated with three skeletal activities are investigated: "Work", "Work Business" and "School". To obtain mutually exclusive homogeneous subsamples, further segmentation of agents is undertaken considering employment and/or student status, and occupation of the individuals. Parametric models are fitted independently to marginal start time and duration distributions using finite mixture models before estimating copula models for their joint distributions. In the "Worker" class, considerable stability is observed in "Work" activity timing models in all occupation sectors for full-time out-of-home employees, except for the "retail sales/service" sector. "Student" class agents' "School" timing behaviour is also found to be stable in general for full-time students. More noise is observed in the timing behaviour of "Both" class agents. Given the correlation levels observed, it is concluded that start time and duration variables should be modelled jointly, not independently, within an agent-based, activity-based travel demand modelling framework. This paper demonstrates that it is possible to capture complex behaviour in timing decisions through parsimonious models that only utilize a small number of parameters and recommends exploiting behavioural information accrued over time while developing models.