Sorry for the slow response. Thanks for the patch! I like the idea and implementation. I do have a few small comments:
1. I am not sure the generification of the confidence interval type is buying anything and in fact I think it is an annoyance / in the way. I get that maybe we would one day want to have Integer there; but to actually do anything with the interval (like compute its width, half-width, etc.) or even to validate its parameters you have to immediately get doubleValue. Number is just too weak a type to be worth much here, IMO. It does not even implement Comparable. So I would favor just using doubles and eliminating the type signature.
2. There is a lot of missing javadoc.
3. We should make it clear what "confidence level" means and how it is measured. I assume it will be between 0 and 1 and for confidence intervals it should be .95 means 95% confidence.
Thanks again for the patch.