Estimate the difference between two population proportions, such as response rates in two groups. The recommended Newcombe interval keeps accurate coverage even with small samples or rates near 0% or 100%. The Wald interval is shown for comparison.
Example
45 of 120 patients improve on treatment A and 30 of 115 on treatment B. The difference is 0.375 − 0.261 = 0.114. The Newcombe 95% interval is about [−0.005, 0.229]. It just includes 0, so the difference is not quite significant at the 5% level, but a benefit of up to 23 percentage points is plausible.