Calculating Confidence Intervals
Confidence intervals provide more information than a P value and do not in the same way tempt readers to make simplified conclusions; ideally it provides the range of effect measures that are “acceptable” given the data. If the null value (RR, IRR = 1; RD, IRD = 0) is not among these, the result is statistically significant, and the no-effect hypothesis is not a likely explanation for the data. Many would find that this description is too vague, but a more precise description would have to take into consideration a number of assumptions that are beyond this text. To calculate confidence limits for the risk ratio or the relative risk (RR) we need to calculate the variance of the log of the Mantel–Haenszel estimated relative risk.