Note on the spatial quantile of a random vector (Q1185564)

Let \(\|\cdot\|_ p\) denote the \(\ell_ p\)-norm on \(\mathbb{R}^ k\), \(1\leq p\leq\infty\). Let \[ \| X\|_{p,\alpha}=\|((| X_ 1|+(2\alpha-1)X_ 1)/2,\dots, (| X_ k|+(2\alpha-1)X_ k)/2)\|_ p, \] where \(X\) is an \(\mathbb{R}^ k\)-valued random variable, and \(0<\alpha<1\). The author defines an \(\alpha\)-quantile of \(X\) as any point \(\theta\in\mathbb{R}^ k\) minimizing the expectation \(E(\| X- \theta\|_{p,\alpha}-\| X\|_{p,\alpha})\). The properties of \(\alpha\)-quantiles and computational aspects are discussed.