ABSTRACT
In Chapter 5, we introduced the multi-category triangular model for a sensitive question with multiple answers based on an assumption that one subclass (e.g., {Y = 1}) is non-sensitive and the other subclasses (denoted by {Y = i} for i = 2, . . . ,m) are sensitive. Thus, the multi-category triangular model cannot be applied to such situations where each subclass is sensitive. Next, the multi-category triangular model still has a lower efficiency for some cases. Third, the parallel model introduced in Chapter 7 can only deal with the case of m = 2 groups where both {Y = 0} and {Y = 1} could be sensitive. These limitations motivate us to introduce a non-randomized multi-category parallel model, which is an extension of the parallel model (Liu & Tian, 2013a).
