Need help with statistics - confidence intervals

[Mark R]

I've got results of a retrospective case-control study in a 2x2 contingency table.

I want to calculate the odds ratio and level of statistical significance. Due to small numbers Fisher's exact test is more appropriate the ?² test.

I've got 2 questions:
1) Is the mid-p statistic more appropriate than the exact statistic? I've received conflicting advice from different sources.
2) I was expecting an odds ratio < 1 - the the exact test yields a one-sided p = 0.04, yet the OR CI is 0.005 - 1.3. What am I interpreting incorrectly?

 

[CollectiveUnconscious]

Originally posted by: Mark R
I've got results of a retrospective case-control study in a 2x2 contingency table.

I want to calculate the odds ratio and level of statistical significance. Due to small numbers Fisher's exact test is more appropriate the ?² test.

I've got 2 questions:
1) Is the mid-p statistic more appropriate than the exact statistic? I've received conflicting advice from different sources.
2) I was expecting an odds ratio < 1 - the the exact test yields a one-sided p = 0.04, yet the OR CI is 0.005 - 1.3. What am I interpreting incorrectly?

As for your first question, mid-p is more appropriate due to your small sample size. It has been growing in acceptance (my entire department uses quasi over exact), and most people recommend that you use some form of quasi-exact alternative. I'll need to think a bit more on question 2.