Understanding Joint Daily Activity Pattern Choices across Household Members Using a Latent Class Model Framework

摘要

In recent years, there has been an emphasis to better understand how intra-householdrelationships impact activity participation decisions. In most operational activity-based model(ABM) systems developed in the U.S., a day activity pattern model is used to describe anindividual’s overall activity engagement for a day. Typically, intra-household interactions arecontrolled for explicitly in such models. This paper uses a latent class model framework of dayactivity pattern to capture these intra-household relationships. This is achieved by defining thelatent class structure at the household level. While independence is assumed across individualactivity pattern choices conditional upon household class membership, the household classstructure, nonetheless, results in interdependencies emerging across household members. Themodel uses data from the Houston-Galveston region and is compared to the day activity patternmodel estimated for that region’s ABM. Results suggest the latent class specification is wellsuitedto this choice context, both in terms fitting the observed data, and in terms of matchingobserved joint distributions of day activity pattern choice across members of the same household.Moreover, the model specification is easy to interpret and implement, and can easilyaccommodate households of any size.