LAUSR

Abstract In this paper a methodology is proposed to consider the importance attributed by users to individual variables in a perceived quality analysis. A two stage ranking based attribute survey is proposed. Firstly, the attributes belonging to each group are ranked and secondly, each group is ranked according to its importance. A series of successive ordered probit models is proposed which also includes models considering systematic and random variations in user taste. The variables are weighted according to the individual and group rankings. The article concludes that increasing the complexity of the models improves their capacity to represent reality, however, there comes a point when the effort required to obtain sufficient data to feed the complexity of the models is not efficient and the time taken is not compensated by the improved predictions.