The application and validation of data fusion

Data fusion is a process used for the merging of separate surveys conducted with different samples of respondents. The result is a single database, en abling the user to understand the relationships between the variables uniqu e to one survey with those unique to the others. Thus, a fused database may purport to explain how the regular readership of a magazine or the audienc e of a TV programme is unusually rich in buyers of a particular grocery ite m, thereby demonstrating the value of including these media in a schedule f or that item. In this paper the authors discuss the rationale for data fusion, applicatio ns of data fusion, and present some empirical results regarding the validit y of data fusion in a particular context. The theoretical assumptions of da ta fusion and the different tests that can be used for validating a specifi c data fusion are also summarised.