0 avis
Assessment of the agreement and cluster analysis of the respondents in a CATA experiment
Archive ouverte
International audience. Statistical tools to assess the agreement of respondents in a Check-All-That-Apply (CATA) experiment are discussed. An overall Index of agreement is introduced and a hypothesis test to assess the significance of this index is outlined. A similar investigation at the level of each attribute is undertaken. The permutation test proposed in this latter situation can be compared to Cochran's Q test. We also propose to cluster the respondents while setting aside those respondents that are deemed to be atypical. This strategy of clustering stands a refinement of the cluster analysis called CLUSCATA that was introduced in a previous paper. The various strategies of analysis are illustrated by means of real case studies.