|
This article is cited in 3 scientific papers (total in 3 papers)
Short Communications
Large deviations of random variables with a finite number of approximately evaluated cumulants
V. I. Bakhtin Belarusian State University, Faculty of Physics
Abstract:
The paper establishes a theorem on exact asymptotics of probabilities of large deviations for random variables with known estimates for only a finite number of cumulants, the latter being subject to conditions of simultaneous growth. For instance, let $S_n$ be a sequence of real random variables and assume the existence of a sequence of small in a sense random variables $G_n(\xi)$ depending on $\xi$ analytically and such that
$$ \mathsf{E}\exp(\xi S_n+G_n(\xi))=\exp\sum_{j=2}^m\frac{\Gamma_{nj}}{j!}\xi^j.
$$
If all the cumulants $\Gamma_{nj}$ have order $n$ and the order of $G_n(\xi)$ does not exceed $n\xi^{m+1}$, then the Cramér type probabilities of large deviations can be indicated for $S_n$.
Keywords:
random variables, distribution function, cumulant, large deviations, Cramer asymptotics.
Received: 26.09.1994 Revised: 13.09.1995
Citation:
V. I. Bakhtin, “Large deviations of random variables with a finite number of approximately evaluated cumulants”, Teor. Veroyatnost. i Primenen., 42:3 (1997), 603–608; Theory Probab. Appl., 42:3 (1998), 513–517
Linking options:
https://www.mathnet.ru/eng/tvp2002https://doi.org/10.4213/tvp2002 https://www.mathnet.ru/eng/tvp/v42/i3/p603
|
|