Calculating an odds ratio with an empty box

Question

How can I calculate an odds ratio with an empty box that would put a zero in the denominator? How can I get a confidence interval?

Mark_Bailey · Answer

I see. I am not familiar with this 'trick,' so I cannot advise you. I imagine, though, that the trick assumes large counts or else it would drastically affect the estimation of the odds ratio and its interval.

KMassa · Answer

Thanks Mark.

I believe I am using JMP 15.

I am using Fit X by Y with categorical variables.  The "positive" outcomes occurs in the intervention group 100% of the time, so I believe that when I ask for an odds ratio, it can't be calculated because the other outcome occurs 0% of the time (and so there is a 0 in the denominator).

I've read about using a small value in place of the 0, in which case I can just calculate an odds ratio by hand, but I'd like to get a 95% CI.  I would think that one end of the 95% CI would be infinite...?

Katherine

Mark_Bailey · Answer

We can help you if you give us more information.

* How are you using JMP?* Which platforms are you using?
* Which models are you using?

* Are you working from a categorical response (data) or do you have a function for the probabilities or odds (model)?
* Are you writing a JMP script?* What kind of empty box are you using?

* What is the scenario in which the denominator would be zero?