On accommodating spatial interactions in a Generalized Heterogeneous Data Model (GHDM) of mixed types of dependent variables

Modeling car-sharing membership as a mobility tool: A multivariate Probit approach with latent variables

“…In this analysis, it is assumed that the choices on mobility tools are made simul- 10 taneously. As such multiple choices may share common underlying unobserved factors and one outcome might be an endogenous factor in another outcome, jointly modeling multiple outcomes accounts for interdependencies and provides deeper insights into the decision making process [4,5,6,7]. 15 For a long time, (motor-)bikes, private vehicles and public transportation subscriptions were the only relevant mobility tools.…”

“…multinomial or ordered outcomes, in which common unobserved factors and endogeneity might be present [6]. The multivariate Probit also allows to accommodate truncated samples [35,36] as well as spatial and social interaction [7].…”

The impact of local work and residential balance on vehicle miles traveled: A new direct approach

“…One can present a list of methodological and behavioral issues that must -in principle -be considered. For example, endogeneity of car ownership in car use (Bhat and Sen, 2006;Bhat et al, 2014), residential self-selection of households with a preference for cars into car friendly locations (Mokhtarian and Cao, 2008;Cao et al, 2009;Ewing and Cervero, 2010) or spatial autocorrelation of observations (Clark, 2007;Bhat et al, 2016), but given the scope of analysis, many of these issues have to be accepted. For example, Cervero and Kockelman (1997) noted that "the relationship between control variables like vehicle ownership and built environment could be endogenous, e.g.…”

Mobility tools and use: Accessibility’s role in Switzerland

“…Joint modeling of multiple related outcomes, e.g., car ownership and use, is motivated by potential common, underlying, unobserved factors in the decision-making process that simultaneously affect outcomes. Ignoring jointness in choices can lead to inefficient estimates of effects and inconsistent estimates of structural effects (Bhat et al 2016 Jointness can be established in several ways. First, multivariate probit-based models consider common underlying factors in multiple outcomes via error term correlation (e.g., Andrés and Gélvez 2014;Scott and Axhausen 2006;Yamamoto 2009).…”

“…Building on the multivariate probit, Bhat and colleagues extended the multivariate probit to model mixed types of dependent variables, e.g., nominal, ordinal, count and continuous outcomes, e.g., location, car ownership, number of trips and trip distance (Bhat 2015;Bhat et al 2014). is modeling approach has also proved suitable for accommodating spatial or social interactions (Bhat et al 2016). Second, copula based models define linking functions between the error terms of outcomes other than the normal distribution, i.e., Gaussian copula (e.g., Spissu et al 2009).…”