33
6
v1v2v3 (latest)

Optimal binomial, Poisson, and normal left-tail domination for sums of nonnegative random variables

Abstract

Let X1,,XnX_1,\dots,X_n be independent nonnegative random variables (r.v.'s), with Sn:=X1++XnS_n:=X_1+\dots+X_n and finite values of si:=EXi2s_i:=E X_i^2 and mi:=EXi>0m_i:=E X_i>0. Exact upper bounds on Ef(Sn)E f(S_n) for all functions ff in a certain class F\mathcal{F} of nonincreasing functions are obtained, in each of the following settings: (i) n,m1,,mn,s1,,snn,m_1,\dots,m_n,s_1,\dots,s_n are fixed; (ii) nn, m:=m1++mnm:=m_1+\dots+m_n, and s:=s1++sns:=s_1+\dots+s_n are fixed; (iii)~only mm and ss are fixed. These upper bounds are of the form Ef(η)E f(\eta) for a certain r.v. η\eta. The r.v. η\eta and the class F\mathcal{F} depend on the choice of one of the three settings. In particular, (m/s)η(m/s)\eta has the binomial distribution with parameters nn and p:=m2/(ns)p:=m^2/(ns) in setting (ii) and the Poisson distribution with parameter λ:=m2/s\lambda:=m^2/s in setting (iii). One can also let η\eta have the normal distribution with mean mm and variance ss in any of these three settings. In each of the settings, the class F\mathcal{F} contains, and is much wider than, the class of all decreasing exponential functions. As corollaries of these results, optimal in a certain sense upper bounds on the left-tail probabilities P(Snx)P(S_n\le x) are presented, for any real xx. In fact, more general settings than the ones described above are considered. Exact upper bounds on the exponential moments Eexp{hSn}E\exp\{hS_n\} for h<0h<0, as well as the corresponding exponential bounds on the left-tail probabilities, were previously obtained by Pinelis and Utev. It is shown that the new bounds on the tails are substantially better.

View on arXiv
Comments on this paper