A serious constraint in a SCE approach to studying preferences in freight is
difficulties in sourcing a large sample size. Indeed, in this application, we
are limited by constraints of both feasibility (that is, the potential to recruit
appropriate respondents) and financial concerns (the cost per computer-aided
personal interview), restricting our expected sample size to 100 respondents or
fewer. As such, we were motivated to extract as much meaningful information
as possible from each respondent, subject to the practicality of time per interview.
Consistent with a previous freight study in Sydney (see Puckett et al, 2007
and Puckett and Rose, 2010 for details), we opted to generate a d-efficient
experimental design to reduce the standard errors associated with attribute
parameters that we would expect to observe in our discrete choice models of
the choice data, relative to standard orthogonal designs for a given sample size