Confidence Intervals for p

By hand, normal approximation method: If npq >= 5, then the sample proportion is approximately normally distributed with expected value p and estimated standard error (sep) = sqrt[(p^q^/n)]. A 95% confidence interval for p is then given by: (sample proportion) ± (1.96)(sep).

When tabulating raw data with the FREQ command, adding the /C option to the command line will calculate 95% confidence intervals for p by an exact method.

When working with summarized data (X and n), use EpiTable | Describe | Proportions | Simple Random Sample to calculate confidence intervals for p. Confidence intervals are calculated by Fleiss's quadratic method, an exact method, and an exact method with mid-p interpolations. The exact method produces the widest (and hence most conservative) confidence interval.