I currently have a sizable dataset containing question-answer variables with multiple levels (4-8) that are unordered and a variable Id like to compare them to, which is a binary variable.

I have looked at a few of the contingency tables, and even the chi-square test statistics for the variables with respect to their relationship to the binary variable, and many of them have "significant" associations.

I was wondering if there was another metric that might help in determining strength of association for these variables, or even a way to see which might be better or "more important" in predicting this variable.

Thanks,

Ross