> +1 on that. I'm an Adjusted Wald fan, personally.
From the JMP docs:
The Confidence Intervals option generates the Score Confidence Intervals table, like the one shown here. The score confidence interval is recommended because it tends to have better coverage probabilities (1 - α), especially with smaller sample sizes (Agresti and Coull 1998).
Not sure if that's the same as the Adjusted Wald calcs. I should go back and read the references I listed! The first ref has some very interesting plots showing actual coverage.
Adj Wald is great for small sample sizes, and converges with most other methods with large ones. I have a couple papers *scanned* that discuss it, but a quick google search has some detail on it: http://www.measuringusability.com/wald.htm