Bounded and compact laws of the logarithm for \(B\)-valued random variables (Q1272164): Difference between revisions

Let \(X\) be a random variable with values in a separable Banach space. The author introduces the following notation: (i) \(X\) satisfies the bounded law of the logarithm (BBL) if for any bounded real-valued field of i.i.d. random variables \(\{W_{n,k}\mid 1\leq k\leq n\); \(n\geq 1\}\) independent of the i.i.d. sequence \(\{X,X_n\mid n\geq 1\},\) \[ \limsup_{n\to\infty}(2n\log n)^{- 1/2}\Big\|\sum_{k=1}^n W_{n,k}X_k\Big\|<\infty, \] (ii) \(X\) satisfies the compact law of the logarithm (CLL) if for any bounded real-valued field of i.i.d. random variables \(\{W_{n,k}\mid 1\leq k\leq n\); \(n\geq 1\}\) independent of the i.i.d. sequence \(\{X,X_n\mid n\geq 1\}\) we have \[ P\Big\{(2n\log n)^{-1/2}\Big\|\sum_{k=1}^n W_{n,k}X_k\Big\| \text{ is conditionally compact}\Big\}=1, \] (iii) \(X\) satisfies the central limit theorem (CLT) if the sequence \(\Big(n^{-1/2}\sum_{k=1}^nX_k\Big)\) converges weakly. The relations between them are studied. In finite-dimensional Banach spaces all these notions coincide and, moreover, are equivalent to e.g. \(E(X)=0\), \(E\big(\| X\|^2\big)<\infty\). Under the additional assumption on \(X\), \(t^2P\big(\| X\|\geq t\big)\to\infty\), necessary and sufficient conditions are given for \(X\) to satisfy BLL (resp. CLL) in terms of moments of \(X.\) In case \(X\) satisfies CLT one equivalence can be expressed simply: \(X\) satisfies CLL iff \(E(\| X\|^2/(1+\log^+\| x\|))<\infty\).