Paired-Comparison Experiments and Logit Models: A Review and Illustration of Some Recent Developments

Publisher Website
Google Scholar
Abstract
Paired-comparison procedures have considerable potential in many areas of geographical research, but in recent years geographers have used them in a rather unimaginative way, confining their use almost exclusively to the initial stages of multidimensional scaling experiments. This paper reviews and illustrates some new procedures for treating paired-comparison data, which link the method of paired comparisons into the rapidly developing unified approach to the analysis of categorical data. The paper concentrates specifically on a weighted-least-squares logit-model approach, and on how to fit such models by use of a readily available computer program, GENCAT.