Short Communications

An error of the Monte-Carlo calculation of the integral by means of a physical generator of random codes

G. A. Kozlov

Leningrad

Abstract: An error of the calculation of a simple integral $\overline\varphi=\int_0^1\varphi\,dx$ by the method of independent tests is estimated in the case when a sequential physical generator of stationary random binary codes with independent digits is used as a source of the random numbers. The imperfection of such a generator can be determined by the value $\varepsilon=P(0)-P(1)$, $P(0)$ and $P(1)$ being the probabilities of 0 and 1 in the code produced.
The error mentioned is estimated by the value
$$ S(v)=\sup\{\Delta\varphi/\sqrt{\mathbf D\varphi}:\ \varphi\in G(v)\}, $$
where $\Delta\varphi=\int_0^1\varphi\,dF-\overline{\varphi}$, $\mathbf D\varphi=\int_0^1(\varphi-\overline{\varphi})^2\,dx$, $F$ is the actual distribution function of random numbers (if $\varepsilon=0$ then $F(x)=x$, $\Delta\varphi=0$ and $S=0$) and $G(v)=\{\varphi:\bigvee_0^1\varphi/\sqrt{\mathbf D\varphi}\le v\}$ is the class of functions with a finite standartized variation.
We prove the relation $\lim_{\varepsilon\to\infty}S(v)/|\,\varepsilon\,|=S^*(v)$ and calculate the function $S^*$. The results may be applied for determining the permissible values of the parameter $\varepsilon$ of the random code generator's imperfection.

Received: 10.05.1976
Revised: 16.04.1978

 English version: Theory of Probability and its Applications, 1981, 25:2, 401–408

Bibliographic databases:

• 
• 
•