Moore-Penrose biorthogonal systems in Euclidean spaces (Q1863575)

A Moore-Penrose generalized inverse of a (real) matrix \(A\) is the (unique) matrix \(A^{+}\) for which \(AA^{+}A=A\), \(A^{+}AA^{+}=A^{+}\), \( (AA^{+})^{T}=AA^{+}\), \((A^{+}A)^{T}=A^{+}A\). A Menger matrix is a square \(n\times n\) real matrix whose entries are squares of the distances of a set of \(n\) points in a point Euclidean space. Using a characterization of the Moore-Penrose generalized inverse of a matrix given by himself and \textit{T. L. Markham} [Linear Algebra Appl. 179, 129-133 (1993; Zbl 0764.15003)], the author defines a Moore-Penrose biorthogonal system of a given system of vectors, pairing in this way two systems of vectors. Then, using some particular simple matrices (\(P=I-(1/n)J\), \(J=ee^{T}\), \(e=(1,\ldots ,1)^{T}\)) and defining the Menger matrix \(M({\mathcal A})\) of a system \({\mathcal A}\) of \(n\) points (using the position vector of the points to the centroid of \({\mathcal A}\)), the author proves that if \({\mathcal A}\) and \({\mathcal B}\) is a Moore-Penrose biorthogonal point pair, then the matrices \((1/2)PM({\mathcal A})P\) and \((1/2)PM( {\mathcal B})P\) are mutual Moore-Penrose inverses. Using some notions from the multivariate statistical analysis [cf. \textit{M. Kendall}, Multivariate analysis (1975; Zbl 0317.62030)], some facts related to a Moore-Penrose biorthogonal point pair are studied in the final part of the paper.