the result is OR= infinity, as in the attached image. So how can report Odds Ratio when one value of a cell is zero?
Hello Islam,
I'd say you have several options:
1. Continue to collect data, as you apparently have insufficient counts to note the presence of any "unexposed" instances among your "cases" batch.
2. Take the cases data (58/0) and replace it with (57/1). The resulting point estimate of the OR would therefore be a slight underestimate of the odds of cases being exposed, or the OR of cases to controls of being exposed, based on your sample data. So you have a lower bound for the point estimate. I'd argue that this is better than, say, replacing the estimated probability of 1.0 for cases being exposed with a value such as 0.999.
3. You could generate a permutation of all possible estimable ORs by splitting the 25 unexposed cases across "cases" and "controls", but that would only yield 24 possible values, so such an approach wouldn't offer a great deal of statistical power, nor would it help with a precise estimate of the population OR.
Good luck with your work.
https://www.jstatsoft.org/article/view/v028i11
https://www.sciencedirect.com/science/article/abs/pii/0167947385900568
In his well-known book Categorical Data Analysis, Alan Agresti recommends adding 0.5 to each cell in the 2x2 table in this situation.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS