Fusion quality assessment

During the past two decades, fusion techniques have been presented and experimented in most European countries, encountering more or less acceptability from the market. Attitudes in front of data resulting from a fusion ranged from indifference or unawareness of users which made no difference in the origin of these data and in the evaluation of their quality or reliability, to a reluctance or even a deny of these techniques. Specially on statisticians side, fusions were not considered as part of the statistical establishment, mainly for validation reasons, as we did not have, for instance, any capability to calculate the confidence intervals around estimates produced by a fusion. Nevertheless, it is a fact that fusion is very efficient to recover information in many cases, and that it helps to answer many questions, as sub-sampling, non-response, length of questionnaire, budget issues, etc… some of these experiences have been presented in past Symposia. This is probably the reason why, a few years ago, fusion started to raise curiosity on the other side of the Atlantic, the program of the forthcoming ARF workshops which dedicates a significant amount of contributions to these techniques is a proof of the interest for fusions among practitioners of market and media research.

Download Paper (PDF)