In an ordered response model the observed variable is based upon classifying an unobserved variable into one out of a finite number of intervals forming a dissection of the real line (cf. Amemiya, 1981). This model considers the boundaries of the intervals as (unknown) deterministic parameters, the same for every individual. Terza (1985) extended this through the relaxation of the assumed constancy of the boundaries: he allowed the boundaries to be a linear function of observed explanatory variables. We extend the deterministic model by allowing for random boundaries that vary across individuals. A case study on consumer valuation of new products indicates that random boundaries significantly improve the standard ordered response model.
|Publication status||Published - 1995|
|Name||CentER Discussion Paper|