Continuous variables for correlations


I hope someone can help me with this question.

If my variable is categorical (1=Sometimes, 2=Frequently ...) can I use it for in a correlational analysis? If not, can I recode my variable into a form such that it can be used?


Hi Justmira,

You can calculate the Spearman's rank correlation coefficient. There are definitions and how-to's if you search on google.


TS Contributor
You can just use the "regular" correlation coefficient (Pearson), but if your variable is a categorical one that represents an increasing/decreasing "quantity" or "frequency," then Spearman's rank-correlation coefficient will provide more meaningful results.

Even when you do an ANOVA, which is usually just for categorical independent variables with continuous dependent variables, the software usually provides a calculation of r and r^2.

There are many different correlation coefficients, depending on the variable types and the levels of measurement:

point biserial -> dichotomous indep variable, continuous dependent variable
phi -> both variables are dichotomies
biserial -> same as point-biserial, but assumes an underlying "continuity" for the dichotomous variable
tetrachoric -> same as phi, but with the assumption of an underlying "continuity"