Pre-limit theorems and their applications (Q1969262)

The authors offer an analytic concept for the following situation what is called pre-limit theorem. In various cases an approximation of a partial sum by a stable random variable is reasonable for intermediate sample size \(n\) although the underlying distribution belongs to the domain of attraction of a normal law and a normal approximation is justitied if \(n\) tends to infinity. The core of the paper is an inequality for a Kolmogorov type metric for distribution functions which is based on a modified probability distance. These inequalities allow a discussion of the Kolmogorov metric for the problem above for intermediate sample size. Reviewer's remark. The same problem appears for binomial random variables of sample size \(n\) and extremly small success probability \(p\). For small and intermediate sample size a Poisson approximation with variational error bound \(np^2\) is reasonable whereas the central limit theorem holds when \(n\) converges to infinity.