I know that to assess the relationship between two categorical variables, you would use a chi square test, but what would you do if you wanted to see the association between a) a ordinal and categorical variable and b) an ordinal and ordinal variable.

I have seen ideas about using gamma, Lambda and taking the linear-by-linear significance, but does anyone have any concrete answers (as this is for my dissertation so I need to make sure I am doing it right).

My data is not normally distributed.

Thank you


1 Answer 1


a) ordinal and categorical variable

Strictly speaking Chi-square test is still a possibility. It tests if the proportion of the attributes in the categorical variables are the same across each rank of the ordinal variable; and it tests if the distribution of the ranks are the same across different attributes in the categorical variables.

However, other tests are available as well. For example, the non-parametric version of ANOVA, called Kruskal-Wallis can let you actually compare the mean ranks of the ordinal variable across attributes of the categorical variable.

b) ordinal and ordinal variable.

A conventional method is Spearman's correlation.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.