Compound Poisson approximations for sums of 1-dependent random variables. I (Q619340)

The paper deals with the (signed) compound Poisson approximations and the second-order asymptotic expansions for certain \(m\)-dependent random variables. The (signed) compound Poisson approximation is obtained as follows. The characteristic function \(\widehat{F}(t)\) of a non-negative random variable is formally written as \[ \widehat{F}(t)=\exp\{\ln\widehat{F}(t)\}= \exp \bigg\{k_1z+\frac{k_2z^2}{2!}+\frac{k_3z^3}{3!}+\cdots\bigg\}. \] The resulting Fourier-Stieltjes transform corresponds to a compound Poisson-structured signed measure. The aim of the paper is to obtain two- or three-parametric compound Poisson approximations for sums of certain 1-dependent random variables. The first part of the paper deals with the simplest case of runs statistics, and the second one considers more general integer-valued random variables that satisfy Franken's condition. A special form of the characteristic function method (Heinrich's method) is used in the proofs. The rate of convergence is estimated in the total-variation and local metrics.