Log Linear analysis is a very useful tool in multivariate set up to analyze the contingency table where the variables are of the qualitative type.

Suppose we have a survey result as below, and we are interested in finding the relationship among sex, heart disease, and body weight.

Typical chi square test for association between variables is insignificant in this case as it tests association among two variables only and can't detect varying interaction (joint effect for two or more variables) and association among the variables more than two.

Log linear analysis is a multivariate extension of the chi square test when one has more than two categorical (or qualitative) variables. The name is derived from the fact that cell frequencies are converted to their natural logarithms for the analysis and the value in a cell is assumed to be a linear function of a set of parameters. This function is called as Log-linear model.