Bug: Prism 6 can mistakenly report that the confidence interval of a median has a 100% confidence level.

When you ask Prism to compute the confidence interval of a median (new to Prism 6), it tries to compute the interval for the confidence level you chose. But given the way the calculations work with ranks, Prism can't always compute that confidence level exactly, so it computes the confidence interval for a similar confidence level and reports the actual confidence level. But in some cases, it mistakenly reports the exact confidence level as 100%, which is of course impossible. 

The bug appears only for small sample sizes. How small depends on the confidence level you requested:

• CI = 99%  : n <= 7
• CI >= 97% : n <= 6
• CI >= 95% : n <= 5
• CI >= 90% : n <= 4
• CI >= 85% : n <= 3
• CI >= 75% : n <= 2

If you see that Prism reported  100% as the "Actual confidence level", the true confidence level is:  

• n = 2 : 50.00%
• n = 3 : 75.00%
• n = 4 : 87.50%
• n = 5 : 93.75%
• n = 6 : 96.88%
• n = 7 : 98.44%

This bug is fixed in Prism 7.