A Model for Ordered Data with Clustering of Observations

Melbourne Institute Working Paper No. 02/00

Date: January 2000


Tim R. L. Fry
Mark N. Harris


There is no qualitative dependent model that can simultaneously account for data sets in which the variable of interest has both a multi-modal distribution and is potentially ordered. Such a multimodal distribution may be the result of individuals being captive to particular choices. Such a case arises when there is digit preferencing (particular numbers, such as 0, 5 and 101 are often favored in many survey-based data sets). This paper introduces a new discrete choice model, the Dogit Ordered Extreme Value (DOGEV), that does account for both ordering and digit preferencing in the data, and applies it to an Australian Inflationary Expectations data set.

Download Paper