Important Stated Preference Experimental Design Issues in Recent Transportation Applications

The paper discusses how the experimental design methods discussed in the preceding chapters, along with other methodological advances, can be used to address areas where SP techniques most need to be improved to deal with particular transportation contexts that are most prevalent in policy applications. Portfolio designs may be useful when people may make a series of choices in a repeating context, such as the mode used to go to work on different days of the week. The paper encourages the use of split sample approaches, where a certain experimental design variation is used for one randomly selected subsample, and one or more other design approaches are tested for other subsamples, allowing the effect of the experimental design on the responses to be tested. This approach can be used for different instructions about how respondents should think about the options, such as in the 'cheap talk' method to avoid non-commitment bias. The split sample approach has also been used to compare different experimental design approaches, such as d-optimal designs versus blocked orthogonal designs. Also encouraged are the use of experimental designs that accommodate more complete model specifications, such as error distributions around parameters, or interaction terms between choice attributes.