topic Re: Correlation Matrix for Categorical Variables in SAS Enterprise Guide

Correlation Matrix for Categorical Variables

gorkemkilic — Wed, 18 Nov 2015 14:53:39 GMT

Hi,

I'd like to see the correlations among the variables in my dataset. I created a correlation matrix for continuous variables by using Analyze/Multivariate/Correlations function but I don't know how to create a correlation matrix for the categorical variables. Could you help me with this issue?

Thanks

Re: Correlation Matrix for Categorical Variables

Reeza — Wed, 18 Nov 2015 17:15:29 GMT

I don't believe that's something that's defined clearly statistically. Typically to compare two categorical variables you can use a Chi-Square or Fisher test.

Re: Correlation Matrix for Categorical Variables

PGStats — Wed, 18 Nov 2015 17:47:04 GMT

"Correlations" are only defined for ordered variables. You can use Spearman rank or Kendall's Tau-b correlations for both continuous measures and ordered categorical variables.

Re: Correlation Matrix for Categorical Variables

ChrisHemedinger — Wed, 18 Nov 2015 17:57:12 GMT

I'm not sure of your application here, but you might look at PROC CORRESP.

PROC CORRESP does analyze categorical data, and it's often used in market research applications, especially in France and Japan. (Which is why I call this the Jerry Lewis of stat procs.)

In fact, there's an entire book about Marketing Research using SAS Enterprise Guide.