A Valuable Addition to Survey Merging Methodologies

Peter Masson
Bucknull and Masson, United Kingdom

Paul Sumner
Bucknull and Masson, United Kingdom


The three methods mentioned all rely on the notion of distance between two informants. A pair of informants is matched when the distance between them is as little as possible. For the ascription methods a distance function is specified as an explicit, calculable quantity for any pair of informants. For calibration the notion of distance is implicit in the cells defined for the calibration.

Several different distance functions (Mahalanobis, Euclidean, Absolute Sum etc.) have been suggested and used for ascription. Experimentation has shown that the particular form of distance function is unimportant. The effect of the particular distance function is completely swamped by the effect of the variables used. We therefore always use the simplest (Absolute Sum) form in our Sesame ascriptions.