1

I have a dataset with several hundred subjects, with rankings in 3 different categories. These rankings are integer values 1-5. My question is whether, for a given subject are these rankings related?

I'm not sure the best way to go about answering this question. Certainly if one particular subject has rankings of 1,1,1 that would support some relationship, but how do I assess this for the whole set of subjects to determine if an association between rankings is likely or not likely to exist?

Any help is appreciated. A similar question has been asked before on this site Aggregation of Correlations Coefficients (Spearman). However, it has not been answered.

kjetil b halvorsen
  • 63,378
  • 26
  • 142
  • 467
scjohnsonwax
  • 11
  • 1
  • 1
    I would think that there is not reason not to extend the Spearman's rank correlation to be three dimensional. It was not developed that way, but so what? – Carl Oct 24 '16 at 20:33
  • I think what you are asking is reasonable, but, it may not ever have been done. So, I asked for rank correlation to be performed for three variables hopefully in a way that will prompt someone to either find an existing solution or create one: http://stats.stackexchange.com/questions/242163/can-spearman-rank-correlation-be-extended-to-three-dimensions – Carl Oct 24 '16 at 21:45

1 Answers1

0

In any real-life population, at least in social science, there's definitely going to be some relationship between the three variables. The world is too messy for anything to come out perfectly independent, or even perfectly uncorrelated.

But we can certainly ask how related the three variables are to each other. If you're interested in monotonic relationships, a good way to do this is to simply compute the three pairwise Kendall correlations.

Kodiologist
  • 19,063
  • 2
  • 36
  • 68