The best constant in the Rosenthal inequality for nonnegative random variables (Q1612944): Difference between revisions

{}t \(X_1,\dots, X_n\) be independent nonnegative random variables with finite \(t\)th moment, \(1\leq t\leq\infty\). H. P. Rosenthal (1970) proved the following important inequality \[ E\Biggl(\sum^n_{k=1} X_k\Biggr)^t\leq A(t)\max\Biggl(\sum^n_{k=1} E(X^t_k), \Biggl(\sum^n_{k=1} E(X_k)\Biggr)^t\Biggr).\tag{Ros70} \] For \(d> 0\), let \(Z(d)\) denote a random variable with Poisson distribution and parameter \(d\), thus \[ P(Z(d)= k)= e^{-d}\cdot d^k/k!,\quad k= 0,1,2,\dots\;.\tag{PoissD} \] Then the main result of the paper is represented by the contents of the following Theorem 1. The best constant \(A^*(t)\) for the Rosenthal inequality (Ros70) is given by the formula \[ A^*(t)= \begin{cases} 2,\quad &\text{for }1< t< 2,\\ E(Z^t(1)),\quad &\text{for }t\geq 2.\end{cases}\tag{BestC} \] .