Suppose I have two assessment methods A and B for assessing depression. I first ask two raters (Raters 1 and 2) to use Method A to assess 50 patients. I compute Cohen's Kappa based on the ratings (assuming the ratings are ordinal). I then ask two different raters (Raters 3 and 4) to use Method B to assess the same 50 patients (again, let's assume the ratings are ordinal). I compute Cohen's Kappa based on these two new raters' ratings.

I wonder if there is a way to compare these two kappa statistics -- that is, whether they are statistically different.


Your best bet would likely be to construct confidence intervals for each kappa point estimate and see if they overlap; if not, then the two estimates are significantly different from one another. Many functions for calculating kappa will include a confidence interval if you ask for it. For instance, in R, you can use irrCAC or agreement to generate Cohen's or Conger's kappa with ordinal weights.

  • $\begingroup$ The link is broken, here are some possible R packages and functions: rel::ckap, irr::kappa2, or psych::cohen.kappa. $\endgroup$
    – Brian D
    Feb 4 at 3:37

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.