I am using ROOT 5.34/09. I have found a strange behavior in the TEfficiency::FeldmanCousins function. For example:

root [0] TEfficiency::FeldmanCousins(1000, 900,0.683,0)
root [1] TEfficiency::FeldmanCousins(10000, 9000,0.683,0)

In the example you see how a factor 10 more statistics produces a much bigger uncertainty.

While the first result agrees with the Clopper-Pearson result, the second one does not.

root [3] TEfficiency::ClopperPearson(1000, 900,0.683,0)
root [4] TEfficiency::ClopperPearson(10000, 9000,0.683,0)

Is this a bug or am I missing anything?


I think it is a bug for large N values. For smaller N seems to work fine. I will investigate it