This formula illustrates the behavior alluded to in the Abstract and in particular proves (3). For p near 0 (resp., 1), Dn(p) is roughly linear in np (resp., n(1 − p)). For p sufficiently far from the endpoints, Sn(p) behaves roughly as √ np(1 − p). Given (1), the only nontrivial relation of the six stated in Theorem 1 is Sn(p) ≤ √ 2Dn(p). 2 The latter is tight (equality is achieved for n = 2, p = 1/2), but the constant √ √ 2 may be improved if we restrict our attention to larger n—but obviously not below π/2, as per (2). In fact, an inspection of the proof of Theorem 1 yields the following: