Interpreting the Parameters of Log-Linear Models

1 August 1987

journal article
research article
Published by SAGE Publications in Sociological Methods & Research

Vol. 16 (1) , 45-77
https://doi.org/10.1177/0049124187016001003

Abstract

This article develops a straightforward approach to the interpretation of the parameters in log-linear models through a detailed consideration of examples. The focus is on models involving dependent variables, and the conventions of regression analysis are used to represent variables in models for the logit, or logarithm of the odds of the dependent variable. By simple manipulations, the parameters in logit models are related to such quantities as odds and odds ratios. The magnitudes of the effects conveyed by the parameters are also described in terms of typical percentage differences. The examples considered include models with complex interaction structures and for polytomous dependent variables. The mechanics of developing parameter estimates from the output of widely available computer programs are shown.

Keywords

This publication has 8 references indexed in Scilit:

Using Adjusted Crosstabulations to Interpret Log-Linear Relationships
American Sociological Review, 1986
Nativity, Intermarriage, and Mother-Tongue Shift
American Sociological Review, 1985
Fitting Nonhierarchical and Association Log-Linear Models Using GLIM
Sociological Methods & Research, 1984
Estimable Functions in Log-Linear Models
Sociological Methods & Research, 1984
Three Parametric Techniques for Contingency Table Analysis: A Nontechnical Commentary
American Sociological Review, 1980
Log-Linear Models for Contingency Table Analysis
Sociological Methods & Research, 1979
Testing for Interaction in Multiple Regression
American Journal of Sociology, 1977
INTERPRETATION OF GOODMAN'S LOG-LINEAR MODEL EFFECTS An Odds Ratio Approach
Sociological Methods & Research, 1977