A note on multivariate stochastic comparisons of Bernoulli random variables (Q1888867)

The following review is a slight modification of the authors' abstract: Let \(\mathbf{I}=(I_1,I_2,\ldots,I_n)\) and \(\mathbf{I}'=(I_1',I_2',\ldots,I_n')\) be two vectors of independent Bernoulli random variables with respective parameters \((p_1,p_2,\ldots,p_n)\) and \((p_1',p_2',\ldots,p_n')\). Let \(X_1,X_2,\ldots,X_n\) be nonnegative exchangeable random variables that are independent of \(\mathbf{I}\) and \(\mathbf{I}'\). The authors show that if \((p_1,p_2,\ldots,p_n)\) is majorized by \((p_1',p_2',\ldots,p_n')\), then \((I_1'X_1,I_2'X_2,\ldots,I_n'X_n)\) is smaller than \((I_1X_1,I_2X_2,\ldots,I_nX_n)\) in the symmetric supermodular stochastic order. The comparison between \((I_1X_1,I_2X_2,\ldots,I_nX_n)\) and \((I_1'X_1,I_2'X_2,\ldots,I_n'X_n)\) in the multivariate symmetric usual stochastic order is also studied. Some results in this paper strengthen results of \textit{C. Ma} [J. Stat. Plann. Inference 84, No. 1--2, 11--25 (2000; Zbl 1131.60303)].