I have a large data set with nominal and dummy variables. What's a good data reduction technique to use? Factor analysis cannot be used here, can it?

Bonus points if there is an R package for this. Extra bonus points if this technique can be used with the survey package. My data is from a complex survey, which I am analyzing using the survey package.

  • 1
    $\begingroup$ Why do you think factor analysis cannot be used? $\endgroup$
    – Gaurav
    Dec 4, 2015 at 9:31
  • $\begingroup$ That's what someone told me in another thread: stats.stackexchange.com/questions/184708/… Apparently, FA needs the data to be continuous, while my data is all nominal variables, which I convert to dummy variables. So is it generally agreed that FA cannot be used with my data? Or is it just the position of some? $\endgroup$
    – Jessica
    Dec 4, 2015 at 13:26

1 Answer 1


If you're willing to assume that some latent categories--as opposed to latent continuous dimensions (as in factor analysis)--underlie your observed variables, you could use latent class analysis using the poLCA package (continuous and categorical predictors are both accommodated). However, I am not sure if poLCA can be used in conjunction with the survey package. Also, poLCA only allows you to model latent categories as outcomes of, as opposed to predictors of, other variables (unlike the latent class analysis capabilities of Mplus).


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.