Hierchical Log-linear Analysis

Abstract

We have already seen in Part 1 of this guide how to examine two-way classifications in contingency tables via the chi-square statistic. As further variables are added to cross-classification tables, the number of cells rapidly increases and it is difficult to unravel the associations between the variables by examining only the cell entries. Although the traditional chi-square approach often provides an insight about the relationships among variables, it:

• Confuses the relationship between a pair of categorical variables with the relationship when other variables are present,

• Does not allow for the simultaneous examination of these pairwise relationships and

• Ignores the possibility of three-variable and higher-order interactions among the variables.