Nonparametric Pooling of Preference Ratings for Conjoint Analysis Experiments with Application on Health Assistance Service

Recent literature on conjoint analysis is rather fragmented and presents some critical elements, both in terms of the procedure for the definition of the survey design and in terms of the subsequent statistical analysis of collected data. The problem of pooling customer preference ratings within a conjoint analysis experiment is addressed. A method based on the nonparametric combination of rankings is proposed to compete with the usual methods based on the arithmetic mean or cluster analysis. The two methods were compared using Spearman’s rank correlation coefficient and a new correlation indicator which takes both correlation and distance between ranks into account. By means of a simulation study it was shown that the nonparametric combination of dependent ranking method performs better than the arithmetic mean under heavy tailed distributions. This aspect is very important, in fact preference ratings in exploratory studies on new product development may largely vary from customer to customer. It is well known that the mean is not a proper indicator for study location under heavy tailed distributions, so the practitioner should take into account the proposed method. Finally, as a real application of the problem of pooling individual preferences, we apply the proposed method to a conjoint analysis case study. Specifically, this study was managed by a health care corporation with the goal of planning the features of a new ambulatory health service. Using a full factorial design, a set of 9 profiles was submitted and rated by a sample of 180 consumers. The study was designed to address the main question on which attributes are essential to consumers, taking also into account the potential segments of consumers, based on their preferences, for the new health assistance attributes.