Approximate independence of distributions on spheres and their stability properties (Q808569)
From MaRDI portal
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | Approximate independence of distributions on spheres and their stability properties |
scientific article |
Statements
Approximate independence of distributions on spheres and their stability properties (English)
0 references
1991
0 references
The authors prove the following result: Suppose \(\zeta_ i\), \(i\geq 1\), is a sequence of independent, identically distributed positive random variables with d.f. F satisfying the normalization E \(\zeta {}_ i^ p=1\) for some \(0<p<\infty\). Define \[ X_{k,n,p}=(\sum^{k}_{j=1}\zeta^ p_ j)/(\sum^{n}_{j=1}\zeta^ p_ j) \] for \(1\leq k\leq n\), \(n\geq 1\). There exists a unique distribution \(F_ p\) with density \[ f_ p(x)=(p^{1-p}/\Gamma (1/p))\exp \{-x^ p/p\} \] such that \(X_{k,n,p}\) has beta-distribution with parameters (k/p, (n-k)/p) for every \(k\leq n\), \(n\geq 1\). Define \[ X_{k,n,\infty}=\max (\zeta_ 1,...,\zeta_ k)/\max (\zeta_ 1,...,\zeta_ n) \] and let \(\gamma_{k,n,\infty}\) have the distribution function \[ P(\gamma_{k,n,\infty}\leq x)=(n-k)n^{-1}x^ k\text{ for } 0\leq x<1,\text{ and } =1\text{ for } x\geq 1. \] Then \(X_{k,n,\infty}\) and \(\gamma_{k,n,\infty}\) are identically distributed for any \(k\leq n\), \(n\geq 1\), iff \(\zeta_ 1\) has uniform distribution on [0,1]. Related characterization theorems and stability properties are discussed. Some discrete versions of the problem and applications to de Finetti-type theorems are investigated.
0 references
approximate independence
0 references
distributions on spheres
0 references
exponential class
0 references
distributions of quotients of sums
0 references
conditional distributions
0 references
maximum entropy condition
0 references
beta-distribution
0 references
characterization theorems
0 references
stability properties
0 references
discrete versions
0 references
Finetti-type theorems
0 references