I have what seemed like a straight-forward analysis to conduct, but perhaps is not. I have two trait categories or sets - behavior, physiology of birds - that I wish to understand the relationship between.

For the behavioral set I've measured three (somewhat related) variables, all quantitative. For the physiological set I've measured four (somewhat related) variables, all quantitative.

What options do I have for reducing each of these two sets to a [single] dimension so that I can correlate the two? The additional challenge is that I do not have a large sample size ($N=25$ individual birds).

  • $\begingroup$ This has been unanswered now for a week. To make somebody helping a little more probable, you should better describe your real problem in the language of the application. So, what is your variables? $\endgroup$ – kjetil b halvorsen Oct 2 '16 at 13:30
  • $\begingroup$ Canonical correlations suggested by mdewey is probably what comes to mind first. It is a multivariate, dimensionality-reducing correlation analysis between sets of variables. $\endgroup$ – ttnphns Oct 2 '16 at 16:41

You might consider canonical correlation described in this Wikipedia article. Basically this would find the linear combination of your behavioural variables and the linear combination of your physiological variables which have maximum correlation. It would show you how your original variables relate to those combinations.

If that does not answer your scientific question you might consider latent variable methods but with that sze of data-set I am not sure you will get much out of them.


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.