Chi-squared tests are used to examine the relationships among categori
cal variables. However, they are difficult to use and interpret when m
ore than 2 variables are involved. In such cases, it is better to use
a related statistic, called log-linear analysis. This article is an in
troduction to log-linear models, illustrating how they can be used to
tease apart relationships among several variables in looking at the fa
ctors associated with photonumerophobia.