“…These various approaches include the use of discrete (see, e.g. Small 1982) versus continuous time models (Wang 1996, van Vuren et al 1999; model structures ranging from logit to other more general models such as nested logit (Brownstone and Small 1989), ordered generalised extreme value (Small 1987), bivariate and multinomial probit (Liu andMahmassani 1998, Lemp et al 2011), error components logit (Bhat 1998, de Jong et al 2003, RAND Europe 2005, Hess et al 2007, Holyoak 2007, Kristoffersson and Engelson 2008, continuous hazard-based logit (Bhat and Steed 2002), continuous cross-nested logit (Lemp et al 2010), and the presence of schedule delay terms, generally available from stated preferences but not revealed preferences surveys (see, e.g. de Jong et al 2003).…”