I work with obesity in cats and many studies use a body condition score (BCS) to assess obesity. This is a somewhat subjective measure of how much fat covering an animal has. There are two commonly used scoring systems for BCS (out of 5 and out of 9).

Does the number of categories have an impact on Spearman correlation between the ordinal ranking out of 5 or 9 and the body fat percent?

EDIT: I am mostly wondering what I should get from the results I am reading. One study will list a Spearman-rank correlation between BCS out of 9 and % body fat of 0.84 with a correlation of 0.72 between weight and % body fat. Another study will list the correlation between BCS out of 5 and % body fat of 0.78 and a correlation of 0.74 between weight and % body fat. Would these differences in correlation with %body fat be expected from the number of categories chosen?

  • $\begingroup$ I don't understand what you are doing, but Pearson/Spearman correlation is used for continuous data, not categorical. Also, merging categories can bias the results. $\endgroup$ – user2974951 Jul 17 '19 at 7:45
  • $\begingroup$ Thank you for the response. I am trying to understand what the differences between the correlations really tells us when they have different category numbers or no categories when directly using weight. So this isn't my data but trying to understand what the data from others really means. I guess my big question is do the above correlations mean that a BCS ranking out of 9 really the best to estimate body fat %. $\endgroup$ – epistudent21 Jul 18 '19 at 14:08
  • $\begingroup$ Yes, splitting data into categories (discretization) will affect measures such as correlation. But as I mentioned, that is not correct, Pearson/Spearman correlation only works for continuous data, so weights in your case. $\endgroup$ – user2974951 Jul 22 '19 at 6:18

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.