Practical Application of Log-Linear Analysis

• In Event: 46.063 - Multiple Regression: Principles and Practices

Abstract

The presentation will discuss a practical way to conduct loglinear analysis. Loglinear analysis is used when analyzing categorical data. The loglinear model is a special type of generalized linear model. Researchers use loglinear models to analyze multiple nested contingency tables that comprise categorical variables.
The presentation will outline the correct steps to use when conducting loglinear analysis. A set of data is utilized for the illustration of the loglinear modeling steps. There are four variables in the data set, namely, gender, region, car, and count. Gender, region, and car are categorical variables, while count represents the cell entry values.
The SPSS example will cover how to weight cases. The illustration will show how to enter the variables in setting up the loglinear model. Other practical issues related to entering the range of the categorical variables, the options available, parameter estimate interpretation, model fit, and association table results will be discussed.
The result interpretation will be explained given that we seek a chi-square value close to 0 (zero). Results will show that the best loglinear equation was: count = car + car*gender + car*region +region*car*gender. Interpretation of the loglinear output will be discussed.
For example, the occurrence of car accidents is a function of gender and region. More specifically, a nested crosstab shows which region and which gender are the main contributors of car accidents.

Author

• Sijia Zhang, The University of Alabama