I keep seeing this in research papers.
The researchers claim that there is a positive correlation between A and B then subsequently show that they odds ratio/sample mean etc. is IN the confidence interval.
How does this make sense? Doesn't that mean that it's exactly NOT statistically significant?
Example:
http://www.ncbi.nlm.nih.gov/pubmed/21972207
There was a positive association between BMI and foot pain (odds ratio [OR] 1.11, 95% confidence interval [95% CI] 1.06-1.17).
So the odds ratio is 1.11 and their confidence interval is 1.06 to 1.17...Which means that the odds ratio is right in the middle of the bell curve, not significant at all??
Am I misunderstanding?
Yes, it seems you are misunderstanding. An odds ratio of 1 implies no effect, i.e., the odds in numerator and denominator are not significantly different. In your example, the confidence interval excludes 1. Hence, on a 5% confidence level, the authors can reject the hypothesis that the odds ratio is 1, i.e., there is no effect.
Moreover, confidence intervals are always constructed around the point estimate (of odds ratio, mean etc.). In that sense the estimate is always in the confidence interval. The relevant research question is whether the prediction of the null hypothesis (no effect) is also in the confidence interval of the point estimate - in the example, it wasn't.