INTERPRETATION OF GOODMAN'S LOG-LINEAR MODEL EFFECTS An Odds Ratio Approach

Publisher Website
Google Scholar
Abstract
In the analysis of multidimensional contingency tables using Goodman's log-linear models, the sometimes complex nature of the problems encountered may lead to an interpretation of analytic results which is based solely upon statistical significance, and not upon substantive importance. In this paper, the well-known relationships between log-linear model effects and odds ratios ("cross-product ratios") are reexamined using previously analyzed sample survey data. The major thrust of the paper concerns the transformation of the log-linear model effects into odds ratios, which then yield simple and direct substantive interpretations.