Second Order Partials as a Means to Control for Diffusion

Abstract

This paper supports Loftin's conclusion that first order partials as cur rently used in linked pair tests are inappropriate. It points out that for each cross-cultural correlation there are not one but at least two control traits: namely one constructed from each substantive variable. This sug gests that not first but second order partials are appropriate diffusion control measures. The paper applies this suggestion to Loftin's hypothetical data and discusses the results.