6533b85cfe1ef96bd12bc839

RESEARCH PRODUCT

Testing the group polarization hypothesis by using logit models

subject

description

This paper focuses on methodological aspects of group polarization research and has two well-defined parts. The first part presents a methodological overview of group polarization research together with an examination of the inadequacy, under certain circumstances, of the traditional parametric approach usually used to test this phenomenon based on pre-test/post-test means comparison across groups. It is shown that this approach will produce masks effects when groups are heterogeneous with regard to the observed change from pre-test to post-test. The second part suggests an alternative methodological approach based on logit models for the analysis of contingency tables from a categorization of the variable ‘kind of shift’. This approach is illustrated and compared with the parametric approach with a simulated data set. Copyright © 2002 John Wiley & Sons, Ltd.