Constructing an Optimal Orthogonal Choice Design with Alternative-specific Attributes for Stated Choice Experiments

1 The stated choice (SC) experiment is generally regarded as an effective way to obtain data for 2 discrete choice analysis. The SC experimental design method, which determines the rule to allocate 3 different levels for each attribute in choice situations, will have a great impact on parameter 4 estimation. The optimal orthogonal choice (OOC) design is one of the most efficient SC designs, by 5 which more reliable parameter estimates can be achieved with an equal or lower sample size. 6 However, OOC design can only be applied to utility models with generic attributes; using it to assign 7 alternative-specific attribute levels is not fully discussed in literature. This paper provides a method 8 to extend the use of OOC design to alternative-specific attributes. Column vectors for 9 alternative-specific attributes are introduced and the value of each vector is forced to be orthogonal 10 with other generic attributes in a same alternative. In this way, orthogonality of OOC design is kept 11 within individual alternative but not necessarily across alternatives. The proposed method was 12 compared with traditional orthogonal design and D-efficient design (another state-of-the-art efficient 13 design method). Three experiments using field data on mode choice were conducted. The result 14 shows that both proposed method and D-efficient design have a higher efficiency than the 15 orthogonal design. In addition, under the complex experiment setting environment in real world, the 16 proposed method outperforms D-efficient design in the sense that almost the same efficiency can be 17 obtained while avoiding multiple iterations for optimal solution. 18 19