Multidimensional limit theorems for large deviations

L. V. Osipov

Teoriâ veroâtnostej i ee primeneniâ, Tome 20 (1975) no. 1, pp. 40-57

Résumé

Let $S_n=X^{(1)}+\dots+X^{(n)}$ be a sum of independent identically distributed random vectors in $R^k$ and let $\Phi$ be the standard normal distribution in $R^k$. Conditions upon distribution of $X^{(1)}$ are given under which $$ \mathbf P\{S_n/\sqrt n\in A_n\}=\Phi(A_n)(1+o(1)),\quad n\to\infty, $$ uniformly in sequences of Borel sets $\{A_n\}$ such that $\Phi(A_n)\ge\Phi(x\colon|x|>\Lambda(n))$ where $\Lambda(z)\uparrow\infty$ is a function satisfying condition (8). In Theorems 1 and 2, we consider the case $\Lambda(z)=bz^\alpha$, $b>0$, $0\alpha1/2$.