JSM 2013- Section on Statistics in Marketing

Fusion and causal analysis in big marketing data sets Since so many marketing studies reflect differing aspects of consumer behavior, there is today a critical need for data fusion. But fusion can be challenging, especially when we are talking about fusing thousands of variables. In this article, we present an approach which solves this problem in a highly efficient manner. Another problem is in applying causal types of models (as opposed to traditional statistical ones) for the analysis of complex marketing data. A new, intrinsic probabilities, approach is proposed and compared with others. 1. Fusion in big data sets The purpose of ascription (fusion) is to merge information of two datasets into one, in a such a way, that external criteria are satisfied. Let assume we have two data sets: A, the source of information to be ascribed (source) and B (hub), with different numbers of respondents.. They both have the same subset of common binary (usually demographic) variables X. Data set A has also