Goodman's Method of Analysis for Multivariate Data — a Practical Application

Abstract

Goodman's method of multivariate analysis is designed for use with nominal and ordinal data, extends to non-recursive models and can explain large variation in the data with residuals greater than 0,9 commonplace. A practical application of this method to a study of factors related to intended emigration by a group of students at the University of the Witwatersrand indicated that the best hypothesis-selected on the basis of a low Chi Square likelihood ratio coupled with a low degree of freedom-explained 97,3% of the data. ‘Religion’ was the variable most directly influentially related to ‘settle’ followed by what students considered the prospects for the future of Whites in South Africa to be, followed by attitudes towards the apartheid policy and ‘religiosity’ in that order. Specific related interlinks could not be omitted from the model, the strongest concerning students who disagreed with the apartheid policy and felt the prospects for the future of Whites were poor.