Properties of Euclidean and non-Euclidean distance matrices (Q1061194): Difference between revisions

A distance matrix D is defined to be a real symmetric matrix with zero diagonal and non-positive off-diagonal entries. D is said to be Euclidean of dimension p if there exist \(x_ 1,...,x_ n\in {\mathbb{R}}^ p\) such that \(D=-(1/2)[\| x_ i-x_ j\|^ 2]\), where \(\|.\|\) is the Euclidean norm, and if p is the smallest dimension for which such \(x_ j's\) exist in \({\mathbb{R}}^ p\). The motivation for studying distance matrices comes from questions of scaling and clustering of multidimensional data. The author discusses the basic mathematical properties of distance matrices, in the belief that his results may eventually help towards a better understanding of statistical methodology, even though they have few immediate statistical applications. In particular he investigates the existence of a circumhypersphere of a collection of points, and establishes a notion of dimension for non-Euclidean distance matrices.