How to interpret infinite odds ratio?

Question

[Lukas]

Hi,

I am currently working on schizophrenia genomic data from case/control study looking for potential patogenic variations. I used these statistics test to check for significant variations:

Fishers exact test - I have only 95 subjects; from scipy Fst statistics - to check how different case from control groups are; I used it because the study have not CS:CN ratio not equal 1:1, but 2:1; from vcftools Odds ratio - to measure association of variation between CS and CN subjects ; from scipy

I have trouble with interpretation with interpretation of Odds ratio inf values. I mistakenly delete them, but i noticed that I have multiple alternatives only in CS group so the inf value is actually division by zero.

What's the best way to interpret these inf odds ratio values accurately?

I am thinking to only extract the inf Odds ratio values to separete table, and check for significance with the Fisher exact test. But it seems to me too simple.

Answer 1

[bhumm]

As you stated, your problem is likely division by 0. See this Cross-Validated post that seems to be very similar to your problem.

In similar situations, I have seen the use of the risk difference to avoid division by 0 problems. See this Cross-Validated post with some potential strategies for your problem: Odds Ratio Vs. Risk Ratio

[Lukas]

Thanks you for your answer. But according that links interpretation of inf odds ratio is up to hypothesis of the researcher.So my solution is actually correct. But there is a big question, the data from disease what I am working at is rare. So the ratio CS:CN was probably done to increase strength of study and I should consider even one heterozygot in case group like positive result.