Paired Interval for Proportions (McNemar)¶
Overview¶
McNemar's test and the associated confidence interval address paired binary data, such as before/after treatment comparisons on the same subjects.
Setup¶
For paired binary outcomes, we construct a \(2 \times 2\) table of concordant and discordant pairs. The CI focuses on the discordant pairs \(b\) and \(c\):
\[
\hat{p}_1 - \hat{p}_2 = \frac{b - c}{n}
\]
McNemar's CI¶
The confidence interval uses the standard error based on discordant pairs.