Ordering distributions on the circle with respect to uniformity (Q915255)

Let \(\mu\) be a probability measure on \(S^ 1\), the circumference of the unit circle. Consider the random variable D(\(\mu\),b) to be equal to the probability mass of the arc \([U,U+b)\), where U has a uniform distribution on \(S^ 1\). Two different notions of variability of probability measures on the circle are studied: \[ \mu_ 1<_ v\mu_ 2\quad if\quad Var(D(\mu_ 1,b))\leq Var(D(\mu_ 2,b)) \] for all \(0<b<1\), and \[ \mu_ 1<_ c\mu_ 2\quad if\quad E f(D(\mu_ 1,b))\leq E f(D(\mu_ 2,b)) \] for all convex functions f and all \(0<b<1\). Properties of these ``orderings'' are explored. For some types of discrete distributions explicit conditions are given which can be used to verify whether or not the orderings hold. In particular empirical distributions are compared. Applications are presented to problems on random coverage, spacings and analysis of directional data.