English

Exact upper and lower bounds on the difference between the arithmetic and geometric means

Probability 2015-07-15 v2

Abstract

Let XX denote a nonnegative random variable with EX<\mathsf{E} X<\infty. Upper and lower bounds on EXexpElnX\mathsf{E} X-\exp\mathsf{E}\ln X are obtained, which are exact, in terms of VXV_X and EXE_X for the upper bound and in terms of VXV_X and FXF_X for the lower bound, where VX:=VarXV_X:=\mathsf{Var}\sqrt X, EX:=E(XmX)2E_X:=\mathsf{E}\big(\sqrt X-\sqrt{m_X}\,\big)^2, FX:=E(MXX)2F_X:=\mathsf{E}\big(\sqrt{M_X}-\sqrt X\,\big)^2, mX:=infSXm_X:=\inf S_X, MX:=supSXM_X:=\sup S_X, and SXS_X is the support set of the distribution of XX. Note that, if XX takes each of distinct real values x1,,xnx_1,\dots,x_n with probability 1/n1/n, then EX\mathsf{E} X and expElnX\exp\mathsf{E}\ln X are, respectively, the arithmetic and geometric means of x1,,xnx_1,\dots,x_n.

Keywords

Cite

@article{arxiv.1503.00345,
  title  = {Exact upper and lower bounds on the difference between the arithmetic and geometric means},
  author = {Iosif Pinelis},
  journal= {arXiv preprint arXiv:1503.00345},
  year   = {2015}
}

Comments

8 pages; to appear in the Bulletin of the Australian Mathematical Society. Version 2: the condition that the random variable X is nonnegative was missing in the abstract

R2 v1 2026-06-22T08:41:11.418Z