ABSTRACT
In this chapter, the authors present extensions of the Warner and unrelated-question models to cover such polychotomous and/or multiattribute situations. Bourke and Dalenius describe a procedure for actually obtaining the data in a polychotomous situation allowing vector responses. The method is essentially based on an extension of the ideas of Morton. Liu and Chow independently describe an elegant randomization device which, incidentally, provides a method of implementing the scheme due to Bourke and Dalenius. The resulting nonlinear programming problem is, therefore, well structured and, as Tamhane indicates, can be solved quite economically on a computer. Tamhane also treats some aspects of the choice of design parameters, keeping in view the question of respondent privacy.
