Abstract: | In the present article, I call for a change in focus in the log-linear analysis of multiway contingency tables. Up to this point, researchers have concentrated largely on procedures for finding the model that most adequately fits the observed data. In many cases, the best fitting model contains one or more complex, higher order interactions involving several variables. It is often desirable to understand the pattern of these interactions. In this article, I present a procedure for substantively interpreting such higher order interactions. This procedure is based on the analogy between log-linear analysis and the general linear model. An example taken from the occupational aspirations literature illustrates the application of the procedure. (PsycINFO Database Record (c) 2010 APA, all rights reserved) |