I've written in last post about two groups of data (case-disease) that need to be compare. This time I don't have binary data but variable data. Example below:

Sample bacteria1 bacteria2 bacteria3
s1 0.5 0.7 0.1
s2 0.03 0.02 0.9
s3 ...

I have 100 samples with disease and 100 samples as control.

To compare the differences between 2 groups (case-control) I've decided to use Mann-Whitney test. But first I would like to predict the sample size for this test with power=0.8 and alpha=0.05. Do you know how can I do that? Is there a tool that can calculate the required sample size for experiment for Mann-Whitney test?

Thanks in advance! Agata

  • 1
    $\begingroup$ Which effect size do you want to detect (in terms of $P(X_{case}>X_{control})$)? $\endgroup$ – Knarpie Mar 1 '17 at 10:30
  • $\begingroup$ Not sure yet, I would like to try different options and fit it to my experiment. $\endgroup$ – Agata Mar 1 '17 at 10:35
  • $\begingroup$ It is only possible to compute sample size once you have defined the minimum effect size you want to detect. It takes huge samples to detect tiny differences; smaller samples to detect huge differences. $\endgroup$ – Harvey Motulsky Mar 1 '17 at 11:36
  • $\begingroup$ Let's say Cohen medium = 0.5 $\endgroup$ – Agata Mar 1 '17 at 11:37

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.