I have a data set which includes 5 binary variables per row of data. I was planning on creating a logistic regression to use 4 of the variables to predict the 5th and measure the significance (if any) of each variable.

Before I do that, I thought I'd look at more simple relationships between the 4 predictors.

What technique can I use to do this? A correlation matrix wouldn't be suitable as they are binary variables.

  • 1
    $\begingroup$ Take a look at Corex, "Correlation Explanation" github.com/gregversteeg/CorEx. The idea is to discover unseen/latent factors that generate your 4 (5?) variables. The number of variables is smallish but can give you insights which group together and with what weight. There might be some hierarchy in the data after all! $\endgroup$ – Vladislavs Dovgalecs Jun 7 '16 at 22:12
  • 2
    $\begingroup$ You can still estimate a correlation matrix. $\endgroup$ – Matthew Gunn Jun 7 '16 at 22:51

Log-linear analysis is a well established, though in some fields largely forgotten, technique for investigating the relationship between multiple categorical variables.

| cite | improve this answer | |

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.