Oscillation measures as randomness tests (Q1204497)

A traditional argument considered by the Italian school of desriptive statistics is the measure of ``oscillation''. Given a sequence of \(n\) numbers \(x_ 1,\dots,x_ n\), the ``oscillation index of order \(h\)'' is defined as: \[ O_ h=\left[(n-1)^{-1}\sum_{i=1}^{n-1}| x_{i+1}-x_ i|^ h\right]^{1/h}, \] where \(h\) is a positive integer. Intuitively, \(O_ h\) is a measure of the ``tendency of \(x_ 1,\dots,x_ n\) to be in a random order''. In fact, \(O_ h\) is miinimum when \(x_ 1\geq x_ 2\geq\dots\geq x_ n\) or when \(x_ 1\leq x_ 2\leq\dots\leq x_ n\), i.e. when the sequence \(x_ 1,\dots,x_ n\) is monotonically ordered. The maximum value of \(O_ h\) has a more complex form. The goal of this note is to study the oscillation indices from an inferential point of view, and to suggest their use as nonparametric test of randomness.