For my experiment, I am comparing the results of melanoma grades( ordinal data, such as Stage 1, 2, 3, 4) as obtained from my proposed noninvasive imaging modality to that obtained from the gold standard (biopsy). In that case, how do I compute the sample size required to have an agreement between these two methods 90% of the times? Is there a specific formula that you know of, since we are comparing ordinal data here?

  • 1
    $\begingroup$ On the one hand, there is likely no calculator out there, so you will need to simulate, which will require that you think about your assumptions (which I consider a feature, not a bug). On the other hand, and relatedly, you will need to assume some baseline level of agreement between your two methods - if they agree only 75% of times (whatever that may mean), then no amount of data will get you to 90% agreement. $\endgroup$ Nov 1, 2023 at 19:30


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.