Designs for stated preference experiments

We explore the use of different strategies for the construction of optimal choice experiments and their impact on the overall efficiency of the resulting design. We then evaluate how these choice designs meet the desired characteristics of optimal choice designs (orthogonality, level balance, utility balance and minimum level overlap). We further explore the feasibility of using entropy as a secondary measure of design optimality. We find that current algorithms afford little flexibility for using this secondary measure. We further study the impact of misspecification of the assumed parameter values used in creation of optimal choice designs. We find that the impact of misspecification varies widely based on the discrepancy between the true and assumed parameter values. Further we find that entropy becomes a more feasible secondary measure of design optimality if one considers the potential of misspecification of the values. Current design and analysis strategies for stated preference experiments assume that compensatory decisions are made. We consider how different decision strategies may be represented through manipulating the assumed parameter values used in creating the choice designs. In this context, the consequences of misspecification of the decision strategy are also evaluated. Given the large prevalence of no-choice choices in stated preference experiments, we study how different measures of choice complexity impact the selection of the no-choice alternative. We conclude by suggesting a comprehensive strategy that should be followed in the creation of choice designs.