I have an analysis question that I need some help on. I received an analysis request to identify events that have statistically different frequencies between two groups.

For example, assume I have group A and group B and in each group, there are events that are either shared or unique to each group. And each event is found to be happening at a certain frequency as denoted in parenthesis below:

Group A: e1 (0.1), e2 (0.03), e3 (0.9), e6 (0.01),e10 (0.02),e11 (0.95)

Group B: e1 (0.2), e2 (0.02), e7 (0.01),e8 (0.02), e12 (0.03)

So the end goal is to identify in this hypothetical case, that event e3 is the significantly different event(occurring at a high frequency in Group A and not at all in Group B) with a p-value and all other events are not significantly different between Group A and Group B. For example, e2 is occurring in both group but at similar frequency.

I was suggested by a friend that binomial model might work but I couldn't figure out how to implement it in R.

Could anyone help me come up with a solution? Thanks!

  • $\begingroup$ Do you have the total number of observations for each group? $\endgroup$
    – J-J-J
    May 20, 2023 at 19:34
  • $\begingroup$ @J-J-J yes, you can add them up in each group to get the total number of events/observations. $\endgroup$ May 20, 2023 at 19:40
  • $\begingroup$ Did you choose event e3 because it had a high frequency in group A and not in group B, or was this a test you wanted to run before you looked at the data? $\endgroup$
    – jbowman
    May 21, 2023 at 2:04
  • $\begingroup$ @jbowman, it's an example of what the test should find out. Yes, it's high freq. in group A and not in group B. $\endgroup$ May 22, 2023 at 2:36
  • $\begingroup$ Would you prefer a test that can simultaneously check all the groups for significant differences? If you just run one test on, in your case, each of12 groups, you will run into the multiple comparisons problem... stats.stackexchange.com/questions/133560/… might be of interest. $\endgroup$
    – jbowman
    May 22, 2023 at 2:50


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.