1
$\begingroup$

We perform data analysis and build models. Say, for example, I built a regression model that has more than one predictor (multiple regression). We then check many things: normality, multicollinearity, etc. Specifically, we check for multicollinearity, for a numeric/continuous variable, VIF (Variance Inflation Factors) etc. If we find that there is multicollinearity, we then drop one of the highly correlated features.

My question is: what can be done with categorical variables? I mean if two categorical variables are correlated/associated does it mean I have to drop it? I am not clear how to handle categorical variables like we handle correlations between continuous variables.

What do we mean by two factor variables are correlated or dependent or independent? What if there is a collinearity? How do you identify that collinearity? How do you deal with it?

$\endgroup$
1
  • $\begingroup$ A category variable could mean categorical-data. 5 categories of specific variable will have to be scored or ranked. The association between two variables can be assessed by chi square test or Pearson test or rank correlation. Correlated here does not mean you should drop both or one of them. $\endgroup$ Apr 10 at 6:10
1
$\begingroup$

Correlation between categorical variables can be calculated with Spearman's rank correlation coefficient. If Spearman's rank correlation is high enough, the variables can be dropped.

$\endgroup$
1
  • $\begingroup$ How would you use that to describe the correlation between pictures of dogs/cats/horses/frogs/BMWs and the picture containing a tennis ball? $\endgroup$
    – Dave
    Jul 13 at 2:56

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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