Visualization Techniques Related to Survey Linking: Two Alternative Ways and Bootstrap Validation.

Abstract

In recent years researchers have focused on investigating the different ways of linking surveys (M. Lejeune, 1995) (L. Lebart & M. Lejeune, 1995b). This is a complex problem that relates the information obtained in two different surveys on the same population. Our main interest is to compare two alternative ways which focus on the relation between two specific sets of variables Y and Z, each corresponding to one of the two surveys considered. The method uses the information from a set of variables X common to the two surveys. We test the robustness of the results using a non-parametric bootstrap, adapting the partial bootstrapping method to the case of survey grafting.