For ν1 = ν2 = 1 this is the non-randomized Clopper–Pearson interval (Clopper and Pearson, 1934) and for ν1 = ν2 = 1/2
this is the non-randomized mid-p interval (Lancaster, 1961). The choice ν1 ∼ U(0, 1) and ν2 = 1 − ν1 corresponds to
the randomized Stevens interval, which is an inversion of the most powerful unbiased test for p (Blank, 1956). We note
that Stevens’ procedure also can be applied when constructing confidence intervals for parameters of Poisson and negative
binomial distributions