Loglinear Analysis

Abstract

Loglinear analysis is a prime example of necessity being the mother of invention. By the early 1970s, the development of multivariate techniques for the analysis of quantitative or interval/ ratio data in the social sciences was well in hand. Analysis of variance, multiple regression and factor analysis, all techniques covered in Modules 6–8 in this text, were firmly established and made up part of the statistical techniques available in SPSS at that time. Most of the variables available in a typical set of social science data, however, are categorical or ordinal variables with a limited number of distinct levels and hence are not suitable for parametric techniques. Social scientists had learned to think about data analysis in terms of multivariate problems but often were frustrated from carrying out multivariate analyses owing to the lack of a technique suitable for nominal or ordinal data. Loglinear analysis was created expressly to fill this gap.