More on the generalized linear model: logistic and loglinear models

Abstract

This chapter discusses sub-frameworks of the generalized linear model (GLZM) which can be used for nominal-level data with two categories (binary logistic model) and scale data in the form of counts (loglinear model). It examines how they can be used to assess the relative effects of multiple explanatory (independent) variables on a single response (dependent) variable. The binary logistic model uses the binomial distribution while the loglinear model uses the Poisson distribution and is useful if the response variable is count data. For explanatory variables, it is the same for the binary logistic and loglinear models as it was for the general linear model: one can have one or more categorical or continuous explanatory variable(s), or indeed a mixture. The chapter then demonstrates how to perform both analyses with SPSS using real data sets, and illustrates how to identify and report the key results of both analyses.